Variation in carbon isotope values among chimpanzee foods at Ngogo, Kibale National Park and Bwindi Impenetrable National Park, Uganda

INTRODUCTION

The application of stable isotope analysis to the study of primate foraging ecology requires an understanding of potential dietary inputs as well as environmentally induced variation. In this article, we focus on quantifying the factors that drive differences in the carbon isotope (δ¹³C) values of plant foods consumed by wild chimpanzees (Pan troglodytes) in two tropical forest habitats in Uganda.

Chimpanzees may serve as an informative model for reconstructing the dietary ecology of extinct apes or early hominins [e.g., see Nelson, 2013]. Wild chimpanzees occupy a range of ecological settings from C₃-dominated closed canopy forest to open woodland and savanna. Given their genetic relatedness to humans and their occupation of contemporary environments perhaps not unlike those utilized by our shared late-Miocene ancestors, analyses of chimpanzee feeding behavior and the isotopic signatures of their habitats may provide insight into reconstructions of late-Miocene forest composition and hominoid behavior. Observed feeding behaviors of chimpanzees vary among sites: fruit consumption ranges from 39 to 64%, pith from 5 to 12%, and leaves from 10 to 30% [as reviewed in Pruetz, 2006]. In the absence of observational data, these dietary differences may be distinguishable using isotopic analyses of hair or bone [Carlson & Kingston, 2014; Schoeninger et al., 1999]. However, in order to make inferences about diet, it is important to first establish that isotopic differences among food sources are consistent among chimpanzee habitats.

Most chimpanzee populations have shown extremely little intra-population isotopic variability [Fahy et al., 2013; Nelson, 2013; Oelze et al., 2014; Schoeninger et al., 1999; Smith et al., 2010; Sponheimer et al., 2006], suggesting that observed differences in feeding behavior are not represented in the carbon isotopic composition of their tissues. Comparing mean δ¹³C values among sites, however, suggests isotopic variability on the order of 3‰ [Cerling et al., 2004; Fahy et al., 2013; Nelson, 2013; Oelze et al., 2014; Schoeninger et al., 1999; Smith et al., 2010; Sponheimer et al., 2006]. It remains unclear if dietary differences, environmental differences, or both are responsible for this variation.

Consistent and predictable isotopic variability across environments is important for ecologically meaningful interpretations using isotopic dietary reconstructions. Rainfall, elevation, and temperature are the primary factors that affect δ¹³C values in C₃ plants at a global scale [Carlson & Kingston, 2014; Crowley et al., 2011; Diefendorf et al., 2010; Kohn, 2010; Körner et al., 1991; Sparks & Ehleringer, 1997]. Additional factors that affect light or water availability at a local or regional scale, such as relative humidity, topography, water use efficiency, rooting depth, and mycorrhizal associations may also impact δ¹³C values in plants [Blumenthal et al., 2012; Broadmeadow et al., 1992; Carlson & Kingston, 2014; Cerling et al., 2004; Cernusak et al., 2009; Codron et al., 2005; Ehleringer & Cooper, 1988; Ehleringer et al., 1986; Farquhar, 1983; Farquhar et al., 1989; Hobbie et al., 1999; Kohn, 2010; Medina & Minchin, 1980; Schoeninger et al., 1998]. Moreover, understory vegetation growing under dense forest canopies typically has lower δ¹³C values than vegetation growing under more open conditions. This is due to lower light levels, higher soil moisture, and the accumulation of ¹³C-depleted CO₂ that is respired by soil microbes [Broadmeadow et al., 1992; Medina & Minchin, 1980; Ometto et al., 2006]. Finally, different parts of the same plant (e.g., leaves vs. fruits) may vary isotopically [Carlson & Kingston, 2014]. This is potentially because of differences in the degree to which these parts rely on sucrose that is fixed during the day versus night, or fractionation associated with translocation of photosynthate from leaves [Codron et al., 2005; Cernusak et al., 2009; Blumenthal et al., 2012].

Stable isotope data for individual food items and plant parts (along with chimpanzee tissue or fecal data) have been reported for several chimpanzee study sites [Blumenthal et al., 2012; Fahy et al., 2013; Oelze et al., 2011, 2014], but comparisons among sites have been hampered by limited sample sizes. As a result, detailed comparisons of chimpanzee isotope ecology across sites remain elusive. We assess the degree to which isotopic variability among known chimpanzee foods (fruit, pith, flowers and leaves) is consistent between two forests in western Uganda: Ngogo (Kibale National Park) and Bwindi Impenetrable National Park. We test the influence of elevation, vertical position in the canopy, plant part, and species on the δ¹³C values of plant samples, and discuss the relevance of our results to dietary reconstructions for living and extinct primates.

METHODS

Plant samples were collected from species consumed by resident chimpanzees at Bwindi Impenetrable (0°53–1°08N, 29°35′–29°50′E) and Kibale (0°13′–0°41′N, 30°19′–30°32′E) National Parks. Both sites are located in Western Uganda along the Albertine Rift (Fig. 1), receive 1,400–1,900 mm of annual precipitation [Chapman & Wrangham, 1993; Eilu & Obua, 2005; Ghiglieri, 1984], and support populations of wild chimpanzees. However, the two parks differ in their forest structure and elevation range. At Kibale, samples were collected from the region occupied by the Ngogo chimpanzees, which is located in the middle of the park and is characterized as a closed canopy lowland tropical rain forest interspersed with open grasslands of Pennisetum purpureum [Struhsaker, 1997, p16]. Elevation at Ngogo ranges from 1,250 to 1,500 m. Bwindi Impenetrable is one of the only forests in Africa with relatively continuous forest cover from 1,200 to 2,600 m (although it is less continuously canopied and has more ground vegetation than Ngogo; Carlson, personal observation). Bwindi has been classified as moist lower montane forest [Chaigneau et al., 2010], and contains some of the highest recorded elevations inhabited by wild chimpanzees. Samples were collected between ca. 1,500 and 2,600 m. We previously found a negative association between elevation and δ¹³C values across the 250 m of elevation change at Ngogo [Carlson & Kingston 2014]. Given its similar rainfall regime but substantial elevation range and different vegetation structure, Bwindi Impenetrable National Park was selected as a suitable location for testing inter-site isotopic variability.

Fieldwork and sample collection was approved by the Ugandan National Council of Science and Technology and the Ugandan Wildlife Authority, and adhered to the American Society of Primatologists principles for the ethical treatment of primates. Data for samples at Kibale are previously published [Carlson & Kingston, 2014]. However, we include details about specimen collection here to illustrate differences in how vegetation samples may be collected and preserved in the field. Plant foods were identified from the literature [Watts et al., 2012], personal observation (BAC), field assistants, and a local expert (Jerry Lwanga). Collection occurred during three weeks in late November and early December 2009 and 8 weeks in June and July 2010. Samples were collected at the approximate stage of maturation most consumed by resident chimpanzees. For example, where ripe fruit and young leaves were preferred, sampling would not include unripe fruit or mature leaves. All foods were collected directly from the tree/plant when accessible without climbing gear (all leaves and pith), or the ground after falling from the canopy naturally (most fruit and flowers). To minimize the potential for contamination or isotopic changes associated with decay, samples with mold, bruising, insect damage, or bite marks were excluded. Samples were dried in the field on the day of collection at 65–77° for up to 3.5 hr using a kerosene fueled stove.

Plant foods consumed by Bwindi chimpanzees were identified from the literature [Stanford & Nkurunungi, 2003] and personal observations of local field assistants. Plant species were identified in the forest by BAC and field assistants, with unknown samples identified or confirmed by a local expert (Robert Barigyira). Samples were collected over the course of two weeks in June 2012 and 1 week in December 2012. Samples were pressed in paper and dried over charcoal stoves. The plant press was placed on a wire rack approximately 1.5 m off the ground. Charcoal stoves (about 30 cm tall) were ignited away from the plant press, and a thick layer of ash poured over the top to quench flames and maintain temperatures below 65°C. Once the stoves began burning without smoke, they were placed under the rack and samples were left to dry overnight.

Isotopic analyses of all plant samples were conducted at the Analytical Chemistry Laboratory in the Odum School of Ecology at the University of Georgia. Samples were run on a Carlo Erba NA-1500 CN Elemental Analyzer coupled to a Thermo Delta V magnetic sector IRMS via a Thermo Conflo III Interface. Analytical error was assessed using an in-house poplar standard for elemental analysis and a bovine collagen standard (NIST 1577-B) for isotopic analysis, with one replicate of each run after every 10 unknown samples. All δ¹³C values are reported in reference to the international V-Pee Dee Belemnite standard.

Statistical analyses were performed using the R program for statistical computing [R Core Team, 2013] and significance was set at α = 0.05. Previously published data from Ngogo [Carlson & Kingston, 2014] were reanalyzed and modeled with the newly collected data from Bwindi. Based on previous results by Carlson and Kingston [2014], multivariate linear models were built stepwise to examine carbon isotopic variation attributable to the following variables: plant part, position within the forest canopy, taxon (defined as species for most samples, and genus for those unidentifiable at the species level), site, and elevation. Because previous analyses at Ngogo indicated no isotopic differences between dry and wet seasons [Carlson & Kingston, 2014], seasonality was not included in these analyses. Additionally, as both parks receive similar amounts of rainfall, precipitation was similarly not included in these analyses.

We divided samples into four “part” categories (fruit, leaves, pith, flowers), which are consumed by chimpanzees at both sites. We also divided samples into two categories based on their canopy position during growth (not necessarily where they were collected): “ground” (0–3 m) and “canopy” (>10 m). Elevation was measured for each sample using a handheld GPS unit. Goodness of fit was assessed by adjusted R-square (herein reported simply as R²) and Akaike Information Criteria (AIC). Analysis of variance (ANOVA) was also used on the full model (with all explanatory variables included) to further examine the degree to which each variable contributes to variation in δ¹³C values. Isotopic differences between sites were assessed using Welch's Two Sample t-tests.

RESULTS

Models

Mean δ¹³C values for chimpanzee plant foods are isotopically equivalent at Ngogo and Bwindi: −28.0 ± 2.8‰ (n = 198) and −27.9 ± 1.9‰ (n = 132), respectively. The range in δ¹³C values is 13.8‰ at Ngogo and 9.1‰ at Bwindi (Fig. 2). Among the taxa with more than one sample, sapling leaves of Morus lactea have the lowest δ¹³C values (−33.7 ± 1.0‰), and Pterygota mildbraedii flowers have the highest values at Ngogo (−24.4 ± 1.0‰) (Table I). At Bwindi, Mimulopsis leaves have the lowest δ¹³C values (−31.3 ± 1.1‰) and Podocarpus milinjianus fruit have the highest values (−24.6 ± 1.1‰) (Table II).

Table I. Summary Carbon Isotope Data for Samples Collected at Ngogo (Reproduced From [Carlson & Kingston, 2014])

Taxon	Canopy location	Food type	n	δ13C Mean ± 1 SD (‰)
Acanthus sp.	Ground	Pith	7	−29.4 ± 2.3
Afromomum sp.	Ground	Pith	8	−29.4 ± 2.3
Aningeria altissima	Canopy	Fruit	5	−25.7 ± 1.5
Celtis africana	Ground	Leaves	9	−28.5 ± 2.4
Celtis durandii	Canopy	Fruit	5	−26.2 ± 0.8
Celtis mildbraedii	Ground	Leaves	3	−30.3 ± 0.8
Chaetacme aristata	Ground	Leaves	5	−28.9 ± 1.8
Chrysophyllum albidum	Canopy	Fruit	9	−27.3 ± 0.9
Cordia millenii	Canopy	Flowers	2	−25.8 ± 1.1
Cordia millenii	Canopy	Fruit	3	−26.4 ± 1.5
Ficus brachylepis	Canopy	Fruit	5	−28.3 ± 1.6
Ficus dawei	Canopy	Fruit	3	−26.0 ± 0.9
Ficus exasperata	Canopy	Fruit	7	−26.3 ± 1.4
Ficus mucuso	Canopy	Fruit	9	−26.9 ± 0.3
Ficus natalensis	Canopy	Fruit	6	−26.9 ± 0.8
Marantochloa sp.	Ground	Pith	5	−32.3 ± 1.7
Mimusops bagshawei	Canopy	Fruit	6	−26.7 ± 0.9
Monodora myristica	Canopy	Flowers	10	−26.6 ± 1.4
Monodora myristica	Canopy	Fruit	8	−27.9 ± 2.3
Monodora myristica	Ground	Leaves	6	−32.1 ± 1.4
Morus lactea	Canopy	Flowers	5	−24.5 ± 0.8
Morus lactea	Canopy	Fruit	5	−25.4 ± 0.5
Morus lactea	Ground	Leaves	9	−33.7 ± 1.0
Neoboutonia macrocalyx	Canopy	Fruit	5	−27.6 ± 0.8
Piper sp.	Ground	Pith	9	−31.3 ± 1.4
Pseudospondias microcarpa	Canopy	Fruit	4	−27.3 ± 0.6
Pterygota mildbraedii	Canopy	Flowers	10	−24.4 ± 1.0
Pterygota mildbraedii	Canopy	Fruit	6	−27.5 ± 1.3
Pterygota mildbraedii	Ground	Leaves	9	−30.9 ± 1.0
Treculia africana	Canopy	Fruit	4	−26.0 ± 1.4
Uvariopsis congensis	Canopy	Fruit	5	−27.4 ± 0.6
Warburgia ugandensis	Canopy	Fruit	6	−24.6 ± 2.2
Total			198	−28.0 ± 2.8

Table II. Summary Carbon Isotope Data for Samples Collected at Bwindi Impenetrable National Park

Taxon	Canopy Location	Food Type	n	δ13C Mean ± 1 SD (‰)
Acanthus sp.	Ground	Pith	2	−26.9 ± 0.7
Aframomum sp.	Ground	Fruit	1	−29.7
Aframomum sp.	Ground	Pith	3	−30.9 ± 0.7
Alchornea hirtella	Ground	Leaves	3	−28.1 ± 0.4
Arundinaria alpina	Ground	Pith	3	−26.2 ± 0.8
Carapa grandiflora	Both	Leaves	2	−28.9 ± 1.0
Cardus sp.	Ground	Flowers	2	−31.2 ± 0.3
Cardus sp.	Ground	Leaves	2	−30.8 ± 0.4
Cassipourea sp.	Ground	Flowers	1	−28.5
Cassipourea sp.	Ground	Leaves	1	−28.9
Chrysophyllum albidum	Canopy	Fruit	1	−25.3
Chrysophyllum gorungosanum	Canopy	Fruit	4	−27.0 ± 0.6
Clematis sp.	Ground	Leaves	3	−28.8 ± 1.5
Dombeya sp.	Ground	Fruit	3	−29.6 ± 0.7
Dombeya sp.	Ground	Leaves	3	−29.6 ± 0.7
Ficus capensis	Canopy	Fruit	2	−26.3 ± 1.0
Ficus natalensis	Canopy	Fruit	3	−28.2 ± 1.4
Ficus pilosula	Ground	Leaves	1	−31.9
Ficus spp.	Both	Fruit	6	−27.8 ± 2.3
Galiniera saxifraga	Ground	Fruit	3	−26.9 ± 0.7
Laplaea mayombensis	Both	Fruit	2	−26.4 ± 2.0
Maesa lanceolata	Both	Fruit	4	−27.3 ± 1.9
Mimulopsis sp.	Ground	Leaves	3	−31.3 ± 1.1
Mormodica calantha	Ground	Leaves	3	−29.7 ± 0.5
Mormodica foetida	Ground	Fruit	3	−29.5 ± 0.6
Mormodica foetida	Ground	Leaves	3	−29.9 ± 0.7
Musanga sp.	Canopy	Fruit	1	−27.1
Myrianthus holstii	Ground	Fruit	2	−29.5 ± 0.0
Mystroxylon aethiopica	Canopy	Fruit	2	−26.6 ± 0.9
Olea capense	Canopy	Fruit	1	−26.1
Olinia usambaransis	Canopy	Fruit	3	−26.1 ± 0.8
Parinari holstii	Canopy	Fruit	2	−28.3 ± 0.2
Piper capense	Ground	Pith	3	−29.2 ± 1.5
Podocarpus millinjianus	Canopy	Fruit	6	−24.6 ± 1.1
Prunus africana	Canopy	Leaves	3	−26.0 ± 0.5
Rhytiginia spp.	Ground	Fruit	3	−27.7 ± 1.8
Rubia cordifolia	Ground	Leaves	1	−29.4
Rubia cordifolia	Ground	Pith	1	−29.2
Rubus sp.	Ground	Fruit	5	−27.4 ± 1.0
Rubus sp.	Ground	Leaves	3	-28.2 ± 0.8
Smilax anceps	Ground	Fruit	2	−26.1 ± 2.3
Strombosia sp.	Canopy	Fruit	2	−26.7 ± 1.5
Symphonia globulifera	Canopy	Flowers	3	−26.8 ± 0.6
Symphonia globulifera	Canopy	Fruit	3	−27.8 ± 0.6
Syzigium spp.	Canopy	Fruit	2	−25.3 ± 0.5
Tabernamontana holstii	Both	Fruit	4	−27.1 ± 2.1
Teclea nobilis	Ground	Fruit	2	−29.1 ± 1.1
Triumphetta sp.	Ground	Leaves	3	−29.0 ± 0.4
Unidentified liana	Ground	Fruit	2	−29.3 ± 1.7
Veronia sp.	Ground	Pith	3	−26.9 ± 0.7
Xymalos monspora	Ground	Leaves	3	−28.6 ± 2.0
Total			132	−27.9 ± 1.9

Four multivariate models produce equivalent R² of approximately 0.70 (Table III). While the full model (δ¹³C as a function of plant part, canopy position, taxon, elevation, and site) produces the highest R² with the lowest AIC value, a reduced model with just canopy position, plant part, and taxon produces a nearly equivalent R² and AIC (Table III). Stepwise model building reveals that canopy position and plant part account for approximately 46% of the total variance in δ¹³C values, with taxon, elevation, and site accounting for an additional 23.7, 6.7, and 4.7% respectively.

Table III. Results for Stepwise Linear Models and Analysis of Variance

Linear models	Adjusted R2	AIC	F-value (df)	P-value
Plant Part + Canopy Position + Taxon + Elevation + Site	0.7059	1,200.6	5.77 (162,160)	<0.0001
Plant Part + Canopy Position + Taxon + Site	0.7059	1,202.4	9.88 (89,240)	<0.0001
Plant Part + Canopy Position + Taxon	0.6992	1,208.7	9.79 (87,242)	<0.0001
Plant Part + Canopy Position + Taxon + Elevation	0.6956	1,211.7	5.63 (159,163)	<0.0001
Plant Part + Canopy Position + Elevation + Site	0.5410	1,265.6	19.97 (20,302)	<0.0001
Plant Part + Canopy Position + Elevation	0.5288	1,266.5	31.12 (12,310)	<0.0001
Plant Part + Canopy Position + Site	0.5095	1,307.1	35.18 (10,319)	<0.0001
Plant Part + Canopy Position	0.4622	1,339.7	46.27 (6,323)	<0.0001
Plant Part + Canopy Position + Elevation + Taxon (Site = Ngogo)	0.7491	717.79	10.15 (62,128)	<0.0001
Plant Part + Canopy Position + Elevation + Taxon (Site = Bwindi)	0.6863	407.29	3.89 (99,32)	<0.0001

ANOVAa	F-value (df)	P-value
Plant Part	3.98 (9)	<0.0001
Canopy Position	58.41 (1)	<0.0001
Taxon	4.25 (64)	<0.0001
Elevation	51.63 (1)	<0.0001
Site	0.46 (1)	0.5
Plant Part * Canopy Position	5.92 (2)	<0.01
Plant Part * Taxon	3.77 (10)	<0.001
Canopy Position * Taxon	3.348 (5)	0.097
Elevation * Taxon	92.48 (57)	0.633

• ^a ANOVA includes all variables: Plant Part + Canopy Position + Taxon + Elevation + Site.

Analysis of variance for the full model reveals significant contributions of plant part, canopy position, taxon, and elevation to the variation in δ¹³C values (Table III). Among the several interactions tested, only plant part + canopy and plant part + taxon are significant (discussed in more detail below). Controlling for other variables in the model, there is no variation in δ¹³C values attributable to site. This result is consistent with equivalent mean δ¹³C values between sites and multivariate linear models, which reveal roughly equivalent goodness of fits with or without study site included (Table III).

Plant Part

There are significant differences in δ¹³C values between sites for leaves, pith, and flowers (Table IV). Pairwise comparisons between plant parts reveal significant differences among most parts at Ngogo, but not at Bwindi (Fig. 3). At both sites, leaves have the lowest δ¹³C values, followed by pith, and fruit (Fig. 3). At Ngogo, flowers have the highest δ¹³C values, whereas at Bwindi, they are intermediate between leaves and pith. While the multivariate linear models and ANOVA both indicate that variation in δ¹³C values is attributable to plant part independently of other variables, there is also a strong interaction between canopy position and plant part. At both sites, all leaf and pith samples were collected at ground level, and most fruit and flower samples came from the canopy (Tables I and II). However, at Bwindi, some sampled flowers (specifically those from the genus Cardus), grow close to the ground, which could lower the mean δ¹³C value for flowers at this site.

DISCUSSION

On average, plants collected at Bwindi and Ngogo are isotopically indistinguishable. However, site averages mask distinctive patterns of isotopic variation between sites. Variability in δ¹³C values for plants at both Ngogo and Bwindi is attributable to vertical position in the canopy, plant part, taxon, and elevation, and the explanatory power of these variables varies between sites. This inter-site isotopic variability has significant implications for reconstructing feeding behavior at sites where the underlying isotopic ecology has not been, or cannot be, defined (e.g., historic or fossil localities).

CONCLUSION

Plant part, taxon, and canopy position all affect δ¹³C values for plants from Ngogo and Bwindi forests in Uganda. Canopy position has the most pronounced effects, and differences in the degree to which canopy and ground resources differ isotopically impact other variables. For example, the reduced isotopic difference between ground and canopy resources at Bwindi reduces isotopic differences among plant parts at this site. Our results indicate that isotopic patterns among food resources at one site may not be applicable at another (even in the same region). This variability between sites in the degree to which food resources differ isotopically has implications for dietary reconstruction efforts in contemporary or past environments. Isotopic differences between populations could result from dietary differences, underlying environmental differences, or both. Analysis of underlying environmental isotopic variation should be included whenever possible. In the absence of this environmental context, it will be challenging to confidently interpret the diet of an individual or population on the basis of tissue isotope δ¹³C values alone.