We are what we eat, plus some per mill: Using stable isotopes to estimate diet composition in Gyps vultures over space and time

Abstract Dietary studies in birds of prey involve direct observation and examination of food remains at resting and nesting sites. Although these methods accurately identify diet in raptors, they are time‐consuming, resource‐intensive, and associated with biases from the feeding ecology of raptors like Gyps vultures. Our study set out to estimate diet composition in Gyps vultures informed by stable isotopes that provide a good representation of assimilated diet from local systems. We hypothesized that differences in Gyps vulture diet composition is a function of sampling location and that these vultures move between Serengeti National Park and Selous Game Reserve to forage. We also theorized that grazing ungulates are the principal items in Gyps vulture diet. Through combined linear and Bayesian modeling, diet derived from δ13C in Gyps vultures consisted of grazing herbivores across sites, with those in Serengeti National Park consuming higher proportions of grazing herbivores (>87%). δ13C differences in vulture feather subsets did not indicate shifts in vulture diet and combined with blood δ13C, vultures fed largely on grazers for ~159 days before they were sampled. Similarly, δ15N values indicated Gyps vultures fed largely on herbivores. δ34S ratios separated where vultures fed when the two sites were compared. δ34S variation in vultures across sites resulted from baseline differences in plant δ34S values, though it is not possible to match δ34S to specific locations. Our findings highlight the relevance of repeated sampling that considers tissues with varying isotopic turnover and emerging Bayesian techniques for dietary studies using stable isotopes. Findings also suggested limited vulture movement between the two local systems. However, more sampling coupled with environmental data is required to fully comprehend this observation and its implications to Gyps vulture ecology and conservation.


| INTRODUC TI ON
Gyps vultures, African white-backed (Gyps africanus), and Rüppell's (Gyps rueppelli) are the most abundant of the six species of vultures found in East Africa (Houston, 1990). The Rüppell's vulture is considerably larger than the African white-backed vulture (~8.5 and ~6 kg, respectively; Houston, 1973). Gyps vultures are obligate scavengers that are entirely dependent on carrion resources (Mundy et al., 1992), and they feed on muscle and viscera from large animal carcasses which make up about 85% of their diet (Houston, 1990). Much of their food supply is made up of animal carcasses that have died from disease or malnutrition rather than predator kills (Houston, 1974(Houston, , 1976.
Vultures contribute to nutrient recycling processes and disease regulation in our ecosystems and yet are among the most threatened taxa of birds (Ogada et al., 2012). Around 70% of vultures and other raptorial birds are categorized as threatened by the IUCN with East African Gyps vultures marked as critically endangered (IUCN, 2017). Declines correlate with increased incidences of poisoning, illegal trade, and loss of habitat for native herbivores which provide carrion for vultures (Ogada et al., 2012). Past telemetry and observational studies in Northern Tanzania (Serengeti-Mara ecosystem), and more recent Ruaha-Katavi and Selous ecosystem in Southern Tanzania, suggest distinctions in home ranges for Gyps and other species of vultures (Bracebridge & Kendall, 2019). North and Southern Tanzanian ecosystems, a product of habitat fragmentation, were noted from early zoological expeditions to have diverse and varied ungulate densities as distinctive features (McNaughton & Nicholas, 1986). These ungulate assemblages play an important role in maintaining vulture populations and make up a significant proportion of vulture food supply (Houston, 1974(Houston, , 1976. Therefore, it is likely that there are differences in how Gyps vultures feed on these assemblages based on location. In dietary analysis studies for birds of prey, estimates are based on the examination of food remains or pellets sampled at nests or resting sites (Donázar et al., 2010;Margalida et al., 2012;Real, 1996). These methods document prey items at high taxonomic resolution (Hidalgo et al., 2005;Milchev et al., 2012). However, vulture species such as those of the Gyps genus may ingest large amounts of meat from animal carcasses contributing less to sampled remains, and sampled remains may not be directly linked to an individual, making it difficult to establish a correlation between ingested biomass and sampled remains (Margalida et al., 2007). Subsequently, biases from this type of dietary analysis linked to sampled remains may be present in quantitative assessments of diet composition in Gyps vultures. Alternatively, intrinsic markers like stable isotopes can provide a good representation of assimilated diet while allowing for documented diet-tissue isotope fractionation (Hobson & Clark, 1992). There are no published diet-tissue fractionation estimates for δ 13 C, δ 15 N, and δ 34 S in Gyps vultures; however, recent developments in stable isotope ecology have enabled imputation of tissue-specific fractionation factors through "SIDER"-a package for use in R (Healy et al., 2018).
Natural differences in stable isotope ratios in animal tissues have broad applications in ecology (Hobson, 1999). Carbon isotope ratios discriminate C3 and C4 photosynthesis in higher plants (δ 13 C = −24‰ to −34‰ and −6‰ to −19‰, respectively; Smith & Epstein, 1971), but is fairly conservative with trophic level, allowing us to estimate the contribution of C3-and C4-based food sources within a consumer's tissues. δ 15 N increases with trophic level since excreted nitrogen is typically depleted in 15 N (DeNiro & Epstein, 1981;Minagawa & Wada, 1984) allowing estimation of an animal's comparative trophic position (Gannes et al., 1998;Vanderklift & Ponsard, 2003). Sulfur isotope ratios (δ 34 S) of animal tissues are generally used to distinguish proximity to the ocean or freshwater systems since water-derived aerosols are typically enriched in 34 S compared with terrestrial sulfur (Newton, 2016). As with δ 13 C, δ 34 S changes little with trophic level (Δ 34 S tissue-diet = +1.2‰ for keratin; Webb et al., 2017) providing a proxy for geolocation of dietary resources.
For this study, we intended to highlight differences in how Gyps vultures utilize ungulate carrion and the relative contribution of ungulate carrion types to Gyps vulture diet as best derived by δ 13 C in Serengeti National Park and Selous Game Reserve. This study was also interested in identifying vulture movement between the two protected areas; δ 34 S a proxy for geolocation can provide an indication of feeding connectivity, as vultures have been observed in past studies to move great distances in search of food (Houston, 1974(Houston, , 1976. To enrich results interpretation, we estimated tissue-specific trophic discrimination factors (TDFs) for African white-backed (AWB) and Rüppell's (RPV) vultures using "SIDER" and sampled blood and feathers from wild Gyps vultures to estimate diet composition derived from δ 13 C and movement to forage from δ 34 S. δ 13 C and δ 34 S analysis enabled us to glean and reconstruct dietary information derived from <64 days past (Kurle et al., 2013) in whole blood, to the time of the latest feather molt ~95 days (Houston, 1975), providing a time series of recent and past diets. environmental data is required to fully comprehend this observation and its implications to Gyps vulture ecology and conservation.

K E Y W O R D S
African white-backed vulture, diet composition, Rüppell's vulture, stable isotopes, trophic discrimination factors

T A X O N O M Y C L A S S I F I C A T I O N
Applied ecology 2 | MATERIAL S AND ME THODS

| Study area description
Tanzania is an East African country with some of the largest protected areas on the African continent; these areas are characterized by high diversity and densities of large mammalian (>5 kg) carnivores and herbivores, their most prominent biological feature (Keast, 1969). About 90 species of large herbivores exist on the African continent (Maglio & Cooke, 1978), with more than 20 species in large and diverse areas such as Kruger, South Africa, and Serengeti-Mara in Tanzania and Kenya (Cumming, 1982). Pioneer studies on feeding patterns of these herbivores in Northern Tanzania noted a graze-to-browse continuum (grazers, mixed feeders, and browsers) among several species shaping animal communities (Lamprey, 1963). This observed resource partitioning played a major role in our study site selection (Figure 1), to assess how Gyps vultures utilize the graze-to-browse continuum.
Located in Northern Tanzania, Serengeti National Park (2.1540°S, 34.6857°E), experiences seasonal inundation with short and long rains from November to February and March to May, respectively (Ogutu et al., 2008). The park is a prominent grazing ecosystem (Fryxell & Sinclair, 1988) and an ecological unit defined by seasonal movement of migratory ungulates. The most numerous of these ungulates include Zebra (Equus quagga), Buffalo (Syncerus caffer), Wildebeest (Connochaetes taurinus), Topi (Damaliscus lunatus), and Thomson's gazelle (Eudorcas thomsonii) (Bell, 1971), which support a large proportion of vulture food supply in the Serengeti (Houston, 1974(Houston, , 1976 (Matzke, 1971). Frequent imposition of watercourses in Selous Game Reserve's Miombo brings about interspersion of vegetation cover, which in turn creates a remarkably similar distribution of animal species and numbers (Matzke, 1971). Contrary to mass ungulate migration in Serengeti National Park, watercourses in the Selous have limited scarcity of pasture and water rendering the need for mass ungulate movements redundant (Matzke, 1971).  (Lamprey, 1963).

| Data collection
Data were collected for a period of 10 months from August 2018 to May 2019, alternating between Selous Game Reserve and Serengeti National Park. We conducted vehicle reconnaissance surveys within the two protected area systems to establish suitable vulture trapping sites. We made noose lines, which are smooth fishing line (1.70 mm thick 300 lb strength) loosely tied into retractable circles along a ~1m parachute rope making a line frame (Watson & Watson, 1985).
Two-line frames were then laid loosely around provisioned or natural bait (where available) and pegged to the ground by 3-inch × 3 mm metal pegs. Traps were set as early as 07:30 h before peak vulture food search effort which ranges between 08:00 and 12:00 h; we then retreated 50 to 60 m from trap sites to observe vulture activity.
Once vultures were noosed by their feet or neck, we rushed to the trap site, covered them with towels, and secured the birds before untying the nooses. We then proceeded to identify the species, age them by plumage, and take weight measurements. This was followed by drawing 0.5 to 1 ml of blood from tarsal veins on F I G U R E 1 Protected areas in Tanzania where vulture and carcass tissue samples were collected their feet using a 23-gauge syringe. The blood was emptied into a labeled vacuum-sealed, red-topped tube, and placed in an Engel freezer (−5°C). Feather molt takes approximately 95 days per cycle and is asymmetric in primary and secondary feathers on the wings of Gyps vultures (Houston, 1975); therefore, we only sampled tail feathers for all individuals caught. A tail feather was cut near the dermis using straight jaw groove joint pliers. The feather was washed with still bottled water and left to dry for 3 min; it was later placed in a labeled A4 envelope. These procedures were repeated for every individual caught in Serengeti National Park: African white-backed vulture (n = 12), Rüppell's vulture (n = 9), and in Selous Game Reserve: African white-backed vultures (n = 5). Feather samples from Rukwa Game Reserve and Ruaha National Park, African white-backed vulture (n = 5) and Hooded vulture (Necrosyrtes monachus) (n = 1), respectively, were provided to us by North Carolina Zoo, USA, working in those areas.

| Stable isotope analysis
We used a Finn pipette to remove approximately 100 μl of vulture whole blood from each of our sample vials; the blood was then emptied into 2 ml Eppendorf microtubes, frozen for 2 h, and freeze-dried. Frozen muscle tissue samples were also freeze-dried.
Approximately 2.5 mg of vulture blood and tissue samples were weighed into 3 × 5 mm tin capsules. Vulture feather samples were cleaned with a 2:1 chloroform: methanol solution in a 100 ml beaker; they were then left to dry on white napkin tissues for 7 min at room temperature. Approximately 1 × 1 cm barb sections were cut from the pennacea proximal and basal ends of feather vanes. Feather barbs weighing ~1.3 mg were weighed into tin capsules as above.

| Statistical analyses
All analyses were carried out using R Statistical software version 4.3.0 (R Core Team, 2020) and RStudio version 1.2.1335 (RStudio . We calculated the absolute difference in δ 13 C, δ 15 N, and δ 34 S between pennacea proximal and basal feather barbs and plotted the data to check for individual-level dietary differences (Figures S1-S3). We ran linear regression models to look at δ 13 C, δ 15 N, and δ 34 S variation within feather barbs by species and sampling location.
We used the "ggplot2" package (Wickham, 2016) to visualize estimated categorized biomass for 23 common ungulates in Serengeti National Park from 12 months of absolute count data and Tanzania Wildlife Research Institute 2018/2019 ungulate census data. The package was also used to visualize δ 13 C, δ 15 N, and δ 34 S ratios of vulture samples across our study areas. We used the package "SIDER" (Healy et al., 2018) to fit a generalized linear phylogenetic regression model to impute AWB/RPV tissue-specific TDF estimates. The response variables were set as δ 13 C or δ 15 N and explanatory variables and feeding ecology (carnivore) and habitat (terrestrial) set as fixed effects. The tissue type, within-species variation (to account for numerous observations in the same species), and phylogeny were set as random effects. The models were fitted using the animal model in the package MCMCglmm with uninformative priors based on course notes within (Hadfield, 2010). MCMC chain convergence diagnostics using the Rubin-Gelman technique (Gelman & Rubin, 1992) and effective sample sizes were automatically done to assess the reliability of estimated TDFs over our four model runs. "SIDER" is unable to estimate tissue-specific TDF for δ 34 S, and therefore, we adopted a fractionation of +1.2‰ ±0.5‰ (Webb et al., 2017).
We used δ 13 C, δ 15 N, and δ 34 S ratios in blood and feathers to parameterize general linear models (GLMs) that determined diet composition and source for Gyps vultures over space and time in our sampled areas. δ 13 C, δ 15 N, δ 34 S ratios as response variables varied as a function of location, vulture species, tissue type, and interaction between tissue type and vulture species. Alternate general linear models that excluded δ 13 C, δ 15 N, δ 34 S ratios in blood were run to compare the robustness of feathers in defining temporal diet variation. Data from Rukwa Game Reserve and Ruaha National Park were excluded from general linear models and all other analyses because we did not have matching blood samples for collected feather samples to make diet comparisons. We used the package "ggfortify" (Tang et al., 2016) to perform general linear model diagnostics, checking for assumptions of homoscedasticity in residuals (Figures S15-S17).
Stable isotope mixing models (SIMMs) were run with the package "MixSIAR" (Stock & Semmens, 2016) in R to determine diet contribution for vultures in Serengeti National Park. Two models were run using three bio tracers (δ 13 C, δ 15 N, and δ 34 S) with one categorical fixed variable either African white-backed (AWB) or Rüppell's vulture (RPV). Error terms, residual error was selected for to account for potential variations in metabolic rates and/or digestibility in the different species of vultures, while process error was not selected-for (Stock & Semmens, 2016). Prey items (herbivore muscle tissue) were combined a priori into browsers and grazers based on their feeding ecology (Phillips et al., 2005). SIMM 1 with δ 13 C, δ 15 N, and δ 34 S ratios in AWB and RPV blood, SIMM 2 with δ 13 C, δ 15 N, and δ 34 S ratios in AWB and RPV feathers were run using the "normal" MCMC parameters and model convergence was assessed using the Gelman-Rubin and Geweke diagnostics. Low muscle tissue sample sizes limited our ability to replicate comparative mixing models to estimate categorized prey item contribution to diet for vultures in Selous Game Reserve.

| Gyps vulture diet composition and vulture forage movement estimates from general linear models
General linear models informed diet composition for Gyps vultures derived from δ 13 C ratios consisted of grazing herbivores with variations in space and time (Figure 2). Serengeti National Park vultures fed on significantly higher proportions of grazing ungulates compared to those in Selous Game Reserve, and over time, there was a significant difference in diet given by δ 13 C ratios in blood and feathers ( Table 3). The average trophic level of prey items fed on by vultures derived from δ 15 N did not vary by species and sampling location ( Figure 3). However, there were significant differences over time for δ 15 N in blood, proximal, and basal feather barbs (Table 4).

| Relative contribution of prey items to
from browsers (Table 7).

Diet in African white-backed and Rüppell's vultures sampled from
Serengeti National Park and Selous Game Reserve consisted of C4 plant grazing herbivores. Serengeti vultures fed on grazing herbivores with higher δ 13 C values compared to those in Selous (Table 3); this was emphasized by stable isotope mixing models that estimated mean grazer contributions of at least 87% to the diet of both Gyps vulture species in Serengeti National Park (Tables 6 and 7). Serengeti National Park represents a surviving member of prominent grazing ecosystems in the world (Fryxell & Sinclair, 1988) whose mammalian biomass comprises 90% grazing ungulates (Bell, 1971). Furthermore, absolute ungulate counts from our transect surveys indicated more grazer abundances compared to browsing and mixed feeding ungulates, establishing the Serengeti as a grazer-dominated ecosystem ( Figure S13). It is highly likely that this grazing abundance and biomass are readily available to Serengeti Gyps vultures and accounted for observed elevated δ 13 C values.
Limiting resources did not permit comparative absolute categorized counts for Selous Game Reserve; however, abundance estimates for the year 2018/2019 acquired from the TAWIRI census database highlighted higher grazer counts compared to other herbivore forage categories ( Figure S14). Diet composition for Selous vultures derived from δ 13 C indicated they fed on prey items that were slightly depleted in carbon compared to vultures in Serengeti (Table 3); however, that difference was within a grazing diet range (−6‰ to −19‰). There was no discernible difference between diet composition for the different species caught; we suspect this is due to similarities in the feeding ecology of Gyps vultures (Houston, 1990).
Temporal vulture diet comparisons for both sites derived from δ 13 C ratios in blood and feathers suggested no change in diet over time and that observed differences between blood and feather barbs from general linear models in Table 3 were representative of δ 13 C tissue-specific fractionation. Furthermore, these differences were within predicted δ 13 C ratio offsets in Gyps vulture blood and feathers (Table 8) and δ 13 C fractionation estimates for the Californian Condor (New World Vulture) whole blood and feathers (Kurle et al., 2013). Therefore, we are certain Gyps vulture diet in the two sites consisted of grazing herbivores over 159 days (combined F I G U R E 2 δ 13 C ratios in vulture tissues across Serengeti National Park (SER) and Selous Game Reserve (SGR) over time

TA B L E 3
General linear model explaining diet composition derived from δ13C as a function of sampling location, vulture species, tissue type, and an interaction between vulture species and tissue type diet-tissue equilibration time for blood and feathers) before the birds were sampled. The average trophic level of prey items in Gyps vulture diet did not vary by site and species, as was expected for African white-backed and Rüppell's vultures that have similar feeding ecology (Houston, 1990). However, observed δ 15 N variations between vulture blood and feathers from results in Table 4

TA B L E 4
General linear model explaining the average trophic level of prey items derived from δ 15 N as a function of sampling location, vulture species, tissue type, and an interaction between vulture species and tissue type F I G U R E 4 δ 34 S ratios in vulture tissues across Serengeti National Park (SER) and Selous Game Reserve (SGR) over time Bottrell, 1997); however, it can also be influenced by wind-blown material and coastal sea spray that can be rained out (Nehlich, 2015).
Biosynthetic pathways in animals bias sulfur isotope selectivity because it is locked up in large amino acids (Griffiths, 1991), causing it to fractionate less when assimilated (+1.2‰ for mammalian keratin and slightly negative for metabolically active tissues; Webb et al., 2017), making δ 34 S a moderately good proxy for geolocation. Our  for this species-specific shift are beyond the scope of this work; however, the difference is likely associated with temporal shifts in diet and space use (Inger & Bearhop, 2008). The ecological significance of such differences seemingly small could for example have implications in more precise fractionation factor estimates used to Ecosystems at the University of Glasgow, Scotland.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no conflict of interest and that the views expressed herein are those of the authors. Funding acquisition (supporting); Project administration (supporting); Supervision (supporting); Writing -original draft (supporting);

AUTH O R CO NTR I B UTI O N S
Writing -review & editing (supporting).