1. Introduction
Tropical forests are complex habitats with unpredictable fruit availability and large intraspecies and intersite variations. The distribution of trees of the same species can fluctuate in space with clumpy, uniform, or random patterns [
1,
2], and fruiting patterns can be synchronous or asynchronous [
3,
4]. Thus, developing effective foraging strategies to survive may affect the behavior of frugivorous species, such as some non-human primates [
5,
6]. Consequently, their seed-dispersal role may be modified, which may affect the regeneration of already threatened forests and, in the medium term, the quality of the primate diet in terms of diversity and abundance. [
7,
8].
During periods of low fruit availability, primates may reduce group size by decreasing food competition (
Pan troglodytes spp. [
9,
10,
11,
12];
Pan paniscus [
9]). They also expand their diet range by consuming fallback foods, i.e., low-quality items that are eaten in large quantity when preferred foods are not available [
13]. For example, fibrous items allow frugivorous primates to maintain stable carbohydrate levels, for example, terrestrial herbaceous vegetation (THV) for chimpanzees (
Pan troglodytes spp. [
9,
12]), bark for orangutans (
Pongo sp. [
14]), or multiple vegetal parts for neotropical primates (
Ateles belzebuth, Lagothrix lagotricha, Cebus apella, and
Alouatta seniculus [
15]). However, due to their lower nutrient quality, fallback foods must be consumed in large quantities to supplement fruit intake, which alters the activity budget by increasing daily feeding and travel time at the expense of resting time (
Pan troglodytes verus [
16];
Alouatta palliata mexicana [
17]). In addition to these seasonal variations, there are interindividual differences in activity budget and energy balance based on sex, age, weight, reproductive status, or dominance rank [
18,
19,
20].
Today, agricultural expansion is one of the major threats to tropical forests, the main habitat of great apes [
21,
22], which often live outside protected areas and may use forest–farm mosaics and human-dominated landscapes for foraging [
23,
24,
25,
26]. Studying energy balance may help us to understand how modified and threatened habitats may affect great apes, especially since these species are all classified as endangered or critically endangered [
27] and have slow life history traits. Great apes have their first offspring between 10 and 16 years old [
28] and long interbirth intervals of 3 to 8 years [
29]. These features make it difficult to rapidly assess population sustainability through censuses and demographic surveys, especially in fast-changing environments.
Crops represent easily digestible and nutritive foods rich in carbohydrates [
30,
31] sparsely dispersed in space and with high seasonality. Besides physiological consequences related to the potential increase in energy intake and food diversity [
32], this direct proximity with humans also affects primate foraging strategies and behavior as they cope with stress [
33,
34] and avoid predation and risks [
35,
36]. In periods of crop maturity, apes living within farm–forest matrices increase their travel time at the cost of resting (
Pan troglodytes verus [
37];
Pongo abelii [
38]), whereas other primates reduce travel time and increase rest time (
Chlorocebus aethiops pygerthrus [
39,
40];
Macaca silenus [
41];
Papio anubis [
42]). Proximity to crops can also encourage chimpanzees to develop nocturnal activities to avoid field guarding (
Pan troglodytes schweinfurthii [
43]).
Chimpanzees, an endangered species [
27,
44], are mainly frugivorous with a flexible diet that highlights fast adaptation through cognitive skills [
45]. Due to the expansion of agriculture, their home range may be close to gardens, providing opportunities to exploit these nutritious resources (see [
46,
47] for reviews), even if this behavior represents a high-risk activity [
36,
48,
49].
Although information regarding energy balance is useful to better understand the threats affecting endangered species, accurate assessment of the energy expended and assimilated is complicated. First, it requires an ethical, non-invasive approach. Capturing and darting individuals represents significant health risks (injury due to fall from a tree, disease transmission, etc.) and may bias physiologic markers due to stress and excessive movements during the capture process. In addition, such invasive methods interfere with the habituation process, monitoring, and well-being of the community [
50,
51,
52]. Second, methods to estimate energy balance are not adapted to remote study sites and inaccessible species (vegetation, topography, etc.) or unhabituated individuals. Despite such difficulties, Pontzer and Wrangham [
53] estimated the energy cost of chimpanzee traveling and climbing, and N’Guessan et al. [
54] highlighted seasonal variations in chimpanzee energy balance by combining direct observations with equations adapted from human studies. However, the literature on wild apes remains relatively sparse.
In this study, we aimed to improve our understanding of how the activity budget and energy balance of wild chimpanzees vary with maize presence and abundance of wild fruits. We hypothesized that chimpanzees would be opportunistic and consume maize because of its spatial and temporal availability (clustered gardens at the boundaries of the protected area with synchronized maize maturity), regardless of fruit availability in the forest. In this case, when maize is available, we expected a lower proportion of wild fruits in the chimpanzees’ diet and thus lower energy gains from forest fruits—a pattern similar to that observed during periods of wild fruit scarcity. By eating more nutritious crops, chimpanzees will also reduce their foraging effort, meaning more rest, less travel, and thus lower energy expenditures—a pattern similar to that observed during periods of high wild fruit availability. Finally, regardless of maize gains and given how chimpanzees exploit wild resources, we expected similar energy balances between maize and non-maize seasons as a result of reduced wild intakes and expenditures in the former case but increasing them in the latter. An alternative hypothesis could be that chimpanzees only use maize as a fallback food when wild fruit availability in the forest is low.
To test this hypothesis, we studied a chimpanzee community (
Pan troglodytes schweinfurthii) living at the northern extremity of Kibale National Park, Sebitoli (Uganda) in an area known as Sebitoli. There, 82% of the chimpanzees’ home range is surrounded by agricultural activities, including subsistence gardens at the direct forest border [
55]. Maize cob is the main and almost only crop item consumed by the chimpanzee community, and farmers usually cultivate it with high seasonality twice a year [
56].
2. Materials and Methods
2.1. Study Site
Kibale National Park (KNP) is a protected area of 795 km
2 located in southwestern Uganda (0° 13′–0° 41′ N; 30° 19′–30° 32′ E) and composed of mature forest, grass lands, swamps, and regenerating forest mosaic. KNP is well-known for its rich diversity of plants and mammals, including more than 1000 individual threatened eastern chimpanzees (
Pan troglodytes schweinfurthii) living in different communities [
57].
A high human population density is present at the edge of the forest (up to 335 inhabitants/km
2 [
58]), as the park is surrounded by tea estates, eucalyptus plantations, and small farms with cash and subsistence crops [
59,
60]. Maize (
Zea mays) is usually cultivated and harvested by farmers twice a year following the rotation of two wet (March–May and September–November) and two dry seasons (December–February and May–August) [
56,
61].
Located in the extreme north of KNP, Sebitoli area, defined as the home range of the Sebitoli chimpanzee community, is a forest patch covering 25 km
2, bisected by a high-traffic national road and contiguous with agriculture on its western, eastern, and northern boundaries [
55,
62,
63]. The Sebitoli area was commercially logged from 1950 to the 1970s, leading to damage of about 50% of the trees; today, degraded or regenerating forests represent 70% of this area, and only 14% represents old-growth forest [
60]. All crop fields are outside the national park, along the northwestern forest border (
Figure 1).
2.2. Sebitoli Chimpanzee Community
The Sebitoli Chimpanzee Project (SCP) started chimpanzee habituation in the Sebitoli area in 2008, and 12 years later, this research team is composed of 25 Ugandan field assistants, eight of whom follow chimpanzees daily, along with researchers and students. The chimpanzee community size is estimated to be 100 individuals, 60 of which are regularly monitored on a 25 km
2 territory across the national road [
63]. Each chimpanzee is identified with a name and a two-letter code, and its age and birth date are estimated or recorded when possible. Age classes (adult, subadult) were defined according to Pontzer and Wrangham [
64]. The sex ratio of known individuals is 1 male for 1.15 females, and more than 25% of the individuals have disabilities [
65]. Having previously assessed the birth date of the infants by direct observations (date of absence of the females and date of return with a newborn), we distinguished lactating and gestating mothers (MO) from non-maternal females (AF), i.e., non-pregnant females or without a dependent infant, by estimating the mean duration of chimpanzee gestation as 32 weeks [
66,
67,
68] and to 5 years for the lactation period [
69].
2.3. Wild Fruit Availability
Between January 2016 and January 2019, temporal wild fruit availability was evaluated by monthly phenology surveys on 10 transects, each 500 m long, distributed through the Sebitoli chimpanzee community home range [
70]. As many as 445 trees from 46 species known to be eaten by chimpanzees according to long-term SCP data were monitored. For each tree, we attributed a score from 0 (no item) to 4 (maximum) to describe the abundance of fruits, leaves, and flowers. We calculated a monthly food availability index (FAI) for wild fruits only adapted from Hockings et al. [
30]:
where G
i is the basal area of the tree, i, and F
i is its abundance score for an item. Some favorite species for chimpanzees were absent from plots or presented a clumpy distribution, such as
Mimusops bagshawei [
55]. We preferred this FAI index, already used in the study area by Bortolamiol et al. [
60], to those including tree density from plots [
70,
71]. We included ripe fruits from all species and unripe fruits from
Ficus sur, Ficus exasperata, Ficus natalensis, and
Mimusops bagshawei, which are known to be consumed when ripe and unripe by the Sebitoli community according to SCP long-term data. Due to missing data, the FAI was not calculated for 3 out of 37 months (April 2016; July 2017; September 2017). We distinguished high fruit availability months (HFA), i.e., months with FAI values greater than or equal to the mean value of the sum of ripe and unripe fruits during the study period, from low fruit availability months (LFA), i.e., months with FAI value less than the mean value.
2.4. Maize Availability
Chimpanzees consume both ripe and unripe maize cobs, as well as maize stems [
43]. We defined the monthly presence (0/1 score) of maize edible by chimpanzees at the north-western border by computing direct observations, informal interviews with farmers, camera trap data over 2016–2019, and, since August 2017, a census of 72 maize gardens (mean size = 1.1 ha) by three SCP field assistants. On average, maize was considered edible from between 10 and 12 weeks after sowing to harvest (up to 27 weeks).
2.5. Monitoring of Individuals
Between January 2016 and January 2019, each day, one chimpanzee of the community was selected and monitored according to Altmann’s focal animal sampling [
72] on a nest-to-nest basis, described as FNN below, usually from 6:00 am to 6:30–7:00 pm. The focal individual was chosen among the better-habituated adults and subadults as soon as individuals present in the party were identified. When possible, we avoided choosing the individuals already selected during the last 4 days of monitoring. If an individual fitting those criteria was observed before 12:00 am, then the observer could start an FNN. The observer recorded each activity and the time spent with a focus on alimentation and movements in trees (see energy sections below) but also detailed the substrate used and its firmness, as well as the health condition of the focal individual.
The focal individual was considered lost after 45 min in the absence of any present evidence of the party followed (paths or vocalizations) [
53]. During feeding sessions of the focal individual, the ingestion frequency of a given item, i.e., the number of fruits/leaves or the length of the stem of the species eaten in one minute, was counted by the observer every 10 min from the beginning to the end of the session to cover all periods of satiety [
53]. Behavior monitoring was associated with spatial monitoring. The position was automatically recorded every 30 s by a GPS Garmin
® (Nanterre, France) 64CS (hereafter called GPS tracks) held by the observer, who, as far as possible, followed the exact chimpanzee paths.
2.6. Energy Expenditures
We decided to approximate the seasonal energy balance with direct field observations of the activity budget. Because of the lack of literature using this methodology on chimpanzees or other great apes, we relied primarily on the approaches of N’Guessan et al. [
54] and Pontzer and Wrangham [
53] to determine expenditures and intakes and for comparison purposes.
The daily energy balance corresponds to the difference between gains provided by the food resources ingested and expenditures lost by the organism. The total daily energy expenditure (
TDEE) is composed of the basal metabolic rate (
BMR), i.e., the energy required to maintain vital functions of the organism, such as breathing or digestion [
73], and the amount of energy required to realize physical activities (
Ei):
We applied a 1.25 factor for gestating females and a 1.5 factor for lactating females to the
TDEE [
18].
The daily
BMR (in kcal) was calculated by using Kleiber’s equation [
74] (used in [
53,
54]) based on the individual body mass,
Mb:
We used body mass values for
P.t. schweinfurthii estimated by Smith and Junger [
75] (used in [
53]): 43 kg for adult males and 36.9 kg for adult females. We added an additional weight of 5 kg for mothers with a dependent infant [
54]. We divided the 150 detailed behaviors observed during FNN into six categories: feeding (F), resting (R), moving in trees (M), and traveling (T). For social activities (SA), we distinguished high social activities (HSA) from low social activities (LSA) [
76,
77]. An ethogram is available in
Table 1. If two activities were simultaneously realized and recorded, we selected the one with the highest value in terms of energy (gain or expenditure).
2.6.1. Daily Traveled Distance and Traveled Energy
The GPS tracks associated with the FNNs were processed with ArcGIS® 10.2.2 using “Elevation Profile” and “ET GeoWizards” add-ins to extract the exact daily length traveled (DLT), which includes slopes.
To calculate the energy expenditures required for traveling, we used Taylor’s equations [
78] based on the theoretical volume of oxygen consumed for walking with a speed,
vT:
The traveling speed on the ground,
vT, was estimated by Hunt [
79], with 0.88 m·s
−1 for adult males, 0.78 m·s
−1 for adult females, and 0.75 m·s
−1 for adult females with a dependent infant. Energy in kcal was calculated by assuming that 1 L of O
2 requires 4.8 kcal to be assimilated [
78]:
2.6.2. Moving in Trees
We distinguished the energy,
EA, required for vertical movements (i.e., ascending or descending, hanging on vertical branches or on a trunk) from the energy,
EM, required for horizontal movements (i.e., hanging or walking horizontally and less than 22.5° inclined branches). Distances were estimated by the observer, referring to the average forelimb length of adult chimpanzees above 50–60 cm [
80] and tree height. Field assistants regularly tested each other to assess their height-estimation accuracy.
Mermier et al. [
81] suggested that the volume of oxygen consumed by humans for ascension is equivalent to a walking speed of 1.9 m·s
−1. This assessment was tested on wild chimpanzees by Pontzer and Wrangham [
53] and used by N’Guessan et al. [
54] in the following equations:
where
vA is the ascending speed in trees for chimpanzees, which was estimated by Pontzer and Wrangham [
53] as 0.5 m·s
−1.
To calculate the energy required for horizontal moves in trees, we applied the same equations as for traveling, considering the same speed.
2.6.3. Other Activities
We used the equation below with energetics coefficients (
D) relative to each activity,
i: 1.25 for resting, 1.38 for feeding and low social activities, and 2.35 for high social activities [
76,
77].
2.7. Uphill Grade Approximation
Since the Sebitoli community ranges in mid-altitude mountains with deep valleys, we sought to predict energy expenditures required for elevation changes and slopes during travel (
Ew). Thus, we selected Bobbert’s equation [
82], which is a logarithmic relationship between
Ew (in cal·kg
−1·min
−1); the travel speed
vT (in m·min
−1); and α, the mean positive slope in degrees (°):
However, because of a different relationship between expenditures, speed, and body mass than in Taylor’s equation [
78] (4), these results were used only for approximation and comparison and were not included in the interseason analyses.
2.8. Energy Gains
We focused our analysis on 13 fruits known to be consumed by the Sebitoli chimpanzee community with nutritional data available for the study site (
Table 2). Fruit collection and the drying process were realized in 2015 by S. Bortolamiol following Rothman et al. [
83], and dried samples were analyzed by S. Ortmann (see
Supplementary Materials). We calculated the caloric gains ingested per FNN as below:
where
j is the fruit species, and DM is the dry matter.
2.9. Statistical Analysis
Following the central limit theorem [
89,
90], we assumed that large samples approximate a normal distribution; otherwise, we evaluated the normality of small samples with the Shapiro–Wilk normality test. We tested the hypothesis of opportunistic maize consumption with a two-sample Student’s t-test to compare wild food availability between maize and non-maize seasons.
Activity budget was considered as a percentage of observation time, focusing on the three main diurnal activities: feed, rest, and travel. Frugivory was considered the percentage of wild fruit ingestion time over total feeding time. Daily length traveled (DLT), total energy expenditures (TDEE), and energy balance were analyzed on an hourly basis. We only selected FNNs with a total duration greater than 6 h, corresponding to a half day of monitoring, to analyze frugivory, activity budget, DLT, and TDEE. Due to the lack of nutrition data, we only selected FNNs of 6 h or more when the 13 food items studied covered at least 80% of feeding time to analyze the ingestion rate, i.e., kcal per minute of feeding, and energy balance.
For each response variable, we built a linear mixed-effect (LMM) model including maize availability (maize, non-maize) and wild fruit availability (HFA, LFA) as main effects and their interaction and individuals as a random effects. All assumptions were validated, except residuals normality for DLT, TDEE, ingestion rate, and energy balance, even after basic transformations of our data (log and Box-Cox). By plotting the model residuals per individual, we found extreme intraindividual variations. We therefore tested the significance of the random effect with the anova() function between the mixed model and the null model (without random effect). We also built several models including sex and/or age class as covariates and compared their Akaike information criterion (AIC). The random effect appeared to be not significant, and the most appropriate model was the null model.
As the three activities (feed, rest, and travel) can be correlated and in order to limit type I errors due to multiple tests, we decided to conduct multiple analyses of variance (two-way MANOVA) to assess variations between seasons [
91]. Homoscedasticity was tested with Levene’s test, covariance homogeneity with a Box’s M test, and multicollinearity by calculating Pearson’s correlation coefficients for pairwise comparisons. We carried out an analysis of variance (two-way ANOVA) as a post hoc test to assess the effect of seasons on each activity. Then, we ran a two-way ANOVA to analyze DLT, TDEE, ingestion rate, and energy balance between seasons. We used the
t2way() function and
mcp2atm() post hoc test from the WRS2 package for a robust ANOVA based on trimmed means [
92,
93,
94]. We used the chi-square test to analyze frugivory among seasons.
Despite our efforts to alternate focal individuals, the FNN sex ratio was rather unbalanced in favor of adult males; thus, we decided to pool male and female data and did not make interindividual comparisons. To understand how individuals contributed to the studied variables, we carried out non-parametric Kruskal–Wallis tests, followed by the Dunn post hoc test with Bonferroni correction between sex–age categories.
A significance level of α = 0.05 was applied, except for the Box’s M test (α = 0.001), and all tests were conducted on R software v.4.0.3 (Vienna, Austria) [
95].
2.10. Ethical Note
Chimpanzees were observed at a distance of 8 m or more without using invasive methods and without any interaction with the researchers or field assistants. We adhered to the research protocol defined by the guidelines of the Uganda Wildlife Authority and approved by the National Museum of Natural History, Paris, France (Memorandum of Understanding MNHN/UWA/Makerere University SJ 445-12).