INTRODUCTION
Resource partitioning among potential competitors takes a central place in ecological research. Grass-eating herbivores have received much attention in these studies because of the high numbers of coexisting species, in particular in Africa (Cromsigt & Olff Reference CROMSIGT and OLFF2006, Prins & Olff Reference PRINS, OLFF, Newbery, Prins and Brown1998, Prins et al. Reference PRINS, DE BOER, VAN OEVEREN, CORREIA, MAFUCA and OLFF2006). Around 100 species of herbivore larger than 2 kg occur in Africa that feed to a greater or lesser extent on grass (Prins & Olff Reference PRINS, OLFF, Newbery, Prins and Brown1998), and various studies have argued that African grazer assemblages are structured by patterns of resource partitioning among potential competitors (Arsenault & Owen-Smith Reference ARSENAULT and OWEN-SMITH2002). MacArthur & Levins (Reference MACARTHUR and LEVINS1967) proposed a simple model to explain species richness as a function of resource availability and niche relationships among species. Based on this niche theory, one might expect the habitat–niche width to decline with an increasing number of sympatric species and decreasing resources. Demment & Van Soest (Reference DEMMENT and VAN SOEST1985) showed that body size of herbivores shows a relationship along a resource axis of grasses having different quality, expressed as fibre contents. Gut capacity increases linearly with body mass (Clauss et al. Reference CLAUSS, SCHWARMM, ORTMANN, STREICH and HUMMEL2007, Demment Reference DEMMENT1982, Demment & Van Soest Reference DEMMENT and VAN SOEST1985), whereas energy requirements (expressed as metabolic body size) scale with actual body weight as W0.75 (Kleiber Reference KLEIBER1975). Since the ratio of metabolic requirements to food processing capacity decreases with increasing size, various authors have argued that smaller-sized herbivores need food of a higher quality than do larger herbivores (Demment & Van Soest Reference DEMMENT and VAN SOEST1985, Jarman Reference JARMAN1974, Prins & Olff Reference PRINS, OLFF, Newbery, Prins and Brown1998). However, the body mass–food quality relationship has recently been challenged. Pérez-Barberia et al. (Reference PÉREZ-BARBERIA, ELSTON, GORDON and ILLIUS2004) could find no significant effect of body mass of ruminants on fibre digestibility, in contrast to previous studies (Iason & Van Wieren Reference IASON, VAN WIEREN, Olff, Brown and Drent1999, Robbins et al. Reference ROBBINS, SPALINGER and VAN HOVEN1995). In addition, Clauss et al. (Reference CLAUSS, SCHWARMM, ORTMANN, STREICH and HUMMEL2007) argued that gut retention time is largely independent of body mass in grazing ruminants, so that a larger body mass does not necessarily lead to improved digestive efficiency.
In a recent study of the ungulate assemblage of Kruger National Park, South Africa, Codron et al. (Reference CODRON, LEE-THORP, SPONHEIMER, CODRON, DE RUITER and BRINK2007) showed a significant negative correlation between body mass and the percentage nitrogen (N) in the faeces of herbivores. In addition, evidence was found of a positive relationship between the body mass of antelopes and buffalo (family Bovidae) and fibre concentrations of the faeces (Codron et al. Reference CODRON, LEE-THORP, SPONHEIMER, CODRON, DE RUITER and BRINK2007). Although these findings are in line with the idea that body mass differences facilitate niche differentiation among sympatric herbivores (the Jarman–Bell principle; Bell Reference BELL1971, Jarman Reference JARMAN1974), no relationship was found between herbivore body mass and diet type (grass versus browse proportions in diet) (Codron et al. Reference CODRON, LEE-THORP, SPONHEIMER, CODRON, DE RUITER and BRINK2007).
Numerous studies have investigated patterns of resource partitioning among grazing herbivores (Cromsigt & Olff Reference CROMSIGT and OLFF2006, Mysterud Reference MYSTERUD2000, Prins et al. Reference PRINS, DE BOER, VAN OEVEREN, CORREIA, MAFUCA and OLFF2006, Putman Reference PUTMAN1996, Voeten & Prins Reference VOETEN and PRINS1999, Wegge et al. Reference WEGGE, SHRESTHA and MOE2006). However, the majority of studies tried to interpret patterns of niche separation based solely on diet composition. Diet can be interpreted as the range of food plants consumed by herbivores. However, the range of food plants found in the diet is not informative about the quality of the food that is consumed. In this study we will test the hypothesis that diet quality will decrease with increasing body mass. We will take an integrative approach to diet composition, niche overlap and food quality to investigate patterns of resource partitioning, and we relate these patterns to the body masses of herbivore species. This is done for an ungulate assemblage in the little-studied Guinea savanna zone of West Africa that is characterized by comparatively nutrient-poor soils and high rainfall.
MATERIALS AND METHODS
Study area
This study was carried out in Bénoué National Park (1800 km2; 7°55′–8°40′N, 13°33′–14°02′E) in north Cameroon. The area has an annual rainfall of 1200−1500 mm, unimodally distributed from April to October. Poor ferruginous soils prevail (Brabant & Humbel Reference BRABANT and HUMBEL1974). The park lies within the northern Guinean savanna zone and is characterized by five major plant communities: Isoberlinia doka woodland savanna, Anogeissus leicocarpus riparian forest, Terminalia laxiflora and T. macroptera open savanna, and Burkea-Detarium open savanna (Stark & Hudson Reference STARK and HUDSON1985). Dominant grasses include species of Andropogon, Hyparrhenia and Loudetia. Bénoué NP is home to 17 species of ungulate (Klop & Van Goethem Reference KLOP and VAN GOETHEM2008; Table 1), which are mostly associated with open wooded savanna and riparian vegetation. By far the most common ungulate species in the Bénoué region is Western kob (Kobus kob), whereas other common species include oribi (Ourebiaourebi), red-flanked duiker (Cephalophus rufilatus) and Western hartebeest (Alcelaphus buselaphus major) (Klop et al. Reference KLOP, VAN GOETHEM and DE IONGH2007, Mayaka Reference MAYAKA2002). This study was conducted in two phenological periods: half-way through the dry season, between 7 February and 18 March (hereafter referred to as dry period), when the grass layer consisted mainly of post-fire regrowth, and early in the rainy season, between 1 and 8 June (hereafter called the wet period) when the grass had reached a height of 15–20 cm.
Table 1. The 17 species of ungulate recorded in Bénoué National Park. Zebu (Bos indicus Linnaeus, 1758; 250 kg) is a domesticated species and not included in the Table. Body mass data are taken from Kingdon (Reference KINGDON1997) and refer to the average weight of male and female, rounded to the nearest kg.
Dung collection and sampling
Diet composition and diet quality of the different herbivores were assessed from dung samples (De Jong et al. Reference DE JONG, GILL, VAN WIEREN and BURLTON1995, Reference DE JONG, VAN WIEREN, GILL and MUNRO2004). Dung samples of seven wild herbivore species (Western kob Kobus kob, Western hartebeest Alcelaphus buselaphus major, Defassa waterbuck Kobus ellipsiprymnus defassa, roan Hippotragus equinus, buffalo Syncerus caffer, hippopotamus Hippopotamus amphibius and one unidentified species) and one domesticated species (Zebu cattle Bos indicus Linn.) were collected in the field during the dry period. A total of 4–11 dung samples per species were collected during the dry season and during the early wet season. Based on the size and shape of the droppings the unidentified species must refer to either oribi, common duiker or red-flanked duiker. In the wet season, dung sample collection was restricted to five species (kob, hartebeest, zebu, roan, hippopotamus). The more limited number of species from which dung samples were obtained in the wet season was due to the difficulties in finding dung in the high grass during the wet season. Dung samples were identified by the researchers themselves and often collected after having seen the animal defecating. Older dung samples were compared with fresh samples that were identified with certainty, based on the size, shape and indentation of the pellets. Samples were then double-checked (blind test) by experienced local trackers and finally checked again, using Stuart & Stuart (Reference STUART and STUART2000). In the dry period dung in the field dried out quickly, so that the structure of the plant material remained intact. Samples were taken from dung that was estimated to be less than 2 wk old. According to Leslie et al. (Reference LESLIE, BROWN and JENKS2007) exposure of faeces to weather or insects does not compromise retention of N for 2–3 wk post defecation. In the wet period dung decay rate was high, so that only fresh dung (up to 3 d old) was collected. All samples were air dried. For each period, the samples were pooled per herbivore species, further dried and sterilized by repeated heating at 100 °C. Two parallel sets of subsamples were then taken, one set for microhistological and another for chemical analysis. According to Van Soest (Reference VAN SOEST1994) air-dried dung can be heated to 100 °C without any significant disturbance of N content measurements by Maillard reactions.
Diet composition
Diet composition was based on microhistological examination of herbivore droppings (De Jong et al. Reference DE JONG, GILL, VAN WIEREN and BURLTON1995, Reference DE JONG, VAN WIEREN, GILL and MUNRO2004). Microhistological examination provides an estimate of the ingested biomass per plant taxon, based on the assumption that the surface/dry matter ratio of leaves is more or less constant in different species (Cid & Brizuela Reference CID and BRIZUELA1990, Sparks & Malechek Reference SPARKS and MALECHEK1968, Stewart Reference STEWART1967). Samples for microhistological analysis were softened and further sterilized by autoclaving with some water for about 8 h at 125 °C, then thoroughly mixed and washed in a Waring Blender, strained over a 0.1-mm plankton sieve and stored in 70% ethanol. From every mixed sample ten random grab samples were examined by light microscopy and measured by using a grid of 0.01-mm2 squares in the microscope eyepiece. At least 100 fragments of cuticle or epidermis were identified by comparison with photomicrographs of epidermal material in a reference collection, as detailed elsewhere (De Jong et al. Reference DE JONG, GILL, VAN WIEREN and BURLTON1995, Reference DE JONG, VAN WIEREN, GILL and MUNRO2004). Over 60 species of food plant, mostly grass species occurring in the region, were sampled for reference slides. Epidermis fragments smaller than 0.01 mm2 were ignored as they do not always carry enough features to make identification conclusive. Epidermis fragments were identified to genus or family level. The abundance of each taxon was calculated as a percentage of the total area of the fragments measured (Alipayo et al. Reference ALIPAYO, VALDEZ, HOLECHECK and CARDENAS1992, Cid & Brizuela Reference CID and BRIZUELA1990, Homolka & Heroldová Reference HOMOLKA and HEROLDOVÁ1992, Putman Reference PUTMAN1984, Sparks & Malechek Reference SPARKS and MALECHEK1968, Stewart Reference STEWART1967).
Diet quality
The quality of the faecal droppings in terms of N and P and fibre contentrations may give indications of the differences in the quality of the plants consumed by the different herbivores. Neutral Detergent Fibre (NDF%) concentration was determined, using 0.5 g of faecal material. NDF is a measure of fibre concentration of the faeces, so high NDF values mean low digestibility of the diet. NDF was determined using the ANKOM filter bag procedure (ANKOM Technology Macedon. NY, USA) with omission of the sodium sulphite and the heat resistant α-amylase. Neutral Detergent Solution was prepared following Goering & Van Soest (Reference GOERING and VAN SOEST1970).
The faecal samples were dried at 100 °C, weighed and burned to determine the inorganic concentration. The N and P concentration of samples was determined by using a modified Kjeldahl destruction. Since all dung samples were pooled per species per period, no measure of variance could be calculated.
Resource overlap
We studied the resource overlap between the different herbivores at two levels, namely: (1) individual taxa (species, genus, or higher level) consumed, and (2) food categories. The latter consisted of six groups: graminoids, graminoid old stems, Caesalpiniaceae leaves, other dicot leaves, Caesalpiniaceae flower buds, and miscellaneous. Resource overlap among the herbivores was quantified using the index of Pianka (Reference PIANKA1973), where complete resource overlap (identical diets) results in a score of 1.0 and a completely different diet would result in a score of 0.0 (Legendre & Legendre Reference LEGENDRE and LEGENDRE1998, Pianka Reference PIANKA1973). To assess whether differences in resource use between two species were significant, we compared the calculated Pianka's index to a reference value for resource overlap based on a randomization procedure of our data. This reference value is the amount of resource overlap that is expected on the basis of chance alone, and was calculated as the mean resource overlap from 5000 randomized matrices of diet composition using the program EcoSim. At the plant taxon level the reference value (mean ± SD) was 0.17 ± 0.09 in the dry season or 0.16 ± 0.05 in the wet season; when the food items were grouped into categories the reference value was 0.40 ± 0.05 in the dry season or 0.27 ± 0.04 in the wet season. Thus, we considered values of Pianka's index lower than the reference values indicative of significantly different diets.
The niche breadth of plant taxa and food categories consumed by a particular herbivore species was assessed using Levins’ measure (Levins Reference LEVINS1968):
where B is the niche breadth and pi is the proportion of plant taxon or food category i in the diet (Levins Reference LEVINS1968). This index was standardized to a scale of 0−1 following Hurlbert (Reference HURLBERT1978):
where n is the total number of taxa or food categories consumed by at least one herbivore species. The relation between body mass and diet composition was investigated using a Mantel test (Legendre & Legendre Reference LEGENDRE and LEGENDRE1998). This test was used to quantify the correlation between a dietary dissimilarity matrix and a body mass dissimilarity matrix. Correlation between herbivore body mass and diet quality, i.e. fibre concentration (expressed as neutral detergent fibre (NDF), i.e. total cell wall concentration), percentage nitrogen and percentage phosphorus in the faeces, was investigated using Pearson correlation coefficients. The observed dietary overlap was also compared to the extent of overlap that could be expected on the basis of chance alone, using a randomization procedure of resource use matrices as described above. The mean observed resource overlap in Bénoué NP was then compared to the mean resource overlap from 5000 randomized matrices.
Deviation from randomness was considered significant when the observed diet overlap fell in the extreme upper or lower tail (each set at 2.5%) of the range of simulated overlap values.
RESULTS
Diet composition
Graminoids formed the major share of the dung samples of all herbivores except the small unidentified herbivore species (Figure 1). During the dry season the proportion of graminoids ranged between 10% (small unidentified species) to 90% (hippopotamus), during the wet season this proportion ranged from 60% (zebu) to 90% (hippotamus). All species but zebu had more graminoids in their dung during the wet season compared with the dry season. However, all species but hartebeest had more graminoids old stems in their dung during the dry season, compared with the wet season. Also in the early wet season very soft grass cuticles were found in all dung samples, especially of Andropogon species, which were less present in the samples from the dry season. Leaf sheath or stem fragments have far less specific patterns than leaves; they were noted as ‘grass’, together with leaf fragments that could not be identified. Thick-walled fragments of dry stems and leaf sheaths which occurred in all faecal samples along with grass fragments of better quality were noted separately as ‘graminoid old stems’ (Figure 1).
Figure 1. Diet composition of herbivores in Bénoué N. P. during the dry (a) and wet (b) season. Plant cuticle other = all other plant cuticles, which were not identified; Caes flower buds = Caesalpiniaceae flower buds; Other dicot leaves = fragments of all other dicotyledon leaves which were not identified; Graminoids = fragments identified as green graminoids; Graminoids old stem = fragments identified as old yellow graminoids.
Besides grasses, most dung samples contained pulpy tissue with epidermis resembling flower buds or fruit pods of Isoberlinia doka, the dominant tree in the region. In addition, in the dry season most samples also contained I. doka leaves. Isoberlinia doka leaves were found most in faeces of migrating domestic cattle, probably from branches that local herdsmen chop off. In the early wet season, I. doka leaves were absent from all dung samples but pods were present, especially in cattle dung. Again, these could have been knocked down by herdsmen. We could not identify any of the seed parts found in the dung. In the dry season, no fruits were present but fleshy flower buds with similar structures were (Figure 1), as trees start flowering in March. Isoberlinia doka buds were the main part of the diet of the unidentified small herbivore species. Based on the high proportion of browse in the diet, this species must have been either common or red-flanked duiker. Oribi, the only other small herbivore in the area, can be ruled out because this species is predominantly a grazer.
Resource partitioning
In the dry season, the diet of the unidentified small herbivore species was different from all other species except waterbuck at both the taxon level and the food category level (Table 1), because of its low grass percentage and high percentage of I. doka buds. Waterbuck also differed from zebu and hippo at the taxon level, but showed no differences at the category level. There were no significant differences between the diets of all other species, at the taxon and the category level in both the dry and wet season (Tables 2 and 3).
Table 2. Resource overlap (expressed as Pianka's index) between species pairs of herbivores at both the food category level (upper right half of matrix) and the grass taxon level (lower left half) in the dry season. The number of samples refers to the number of pellet groups that were aggregated to one mixed sample for each species. * indicates significantly different diets. The number of samples refers to the number of pellet groups that were aggregated to one mixed sample for each species.
Table 3. Resource overlap (expressed as Pianka's index) between species pairs of herbivores at both the food category level (upper right half of matrix) and the grass taxon level (lower left half) in the wet season. For details see Table 1.
The niche breadth for food categories consumed by kob (0.300), hippo (0.090), hartebeest (0.350), roan (0.510) and zebu (0.300) was much greater in the dry season than in the wet season for kob (0.120), hippo (0.020), hartebeest (0.190), roan (0.090) and zebu (0.200) (Figure 2a). When looking at grass taxa consumed, the niche breadth of kob (0.220), hartebeest (0.140) and roan (0.250) was also greater in the dry season when compared with the wet season for kob (0.050), hartebeest (0.120) and roan (0.120). The opposite was found for zebu and hippo, with a smaller niche breath for zebu (0.100) and hippo (0.060) in the dry season compared with the wet season for zebu (0.200) and hippo (0.210) (Figure 2b).
Figure 2. Seasonal standardized niche breadth of herbivores in Bénoué N. P., represented by Levins’ niche breadth index, standardized to a scale of 0−1 following Hurlbert (Reference HURLBERT1978). Niche breadth for food categories (a), niche breadth for grass species consumed (b).
A Mantel test showed that there was no significant correlation between body mass and (differences in) diet composition, either for the dry season (taxon level: Z = 0.19, P = 0.72; category level: Z = 0.15, P = 0.66) or the wet season (taxon level: Z = 0.41, P = 0.81; category level: Z = 0.43, P = 0.82). Omitting the unidentified small herbivore species (i.e. common or red-flanked duiker) from the dry season analyses does not change this outcome (taxon level: Z = −0.23, P = 0.82, category level: Z = 0.11, P = 0.37). When hippopotamus, the only non-ruminant in this study, is also omitted from the analyses the relation between body mass and diet composition remains non-significant (dry season, taxon level: Z = −0.28, P = 0.73, category level: Z = −0.19, P = 0.59).
The observed dietary overlap in Bénoué NP was significantly greater than the mean overlap of randomized resource use matrices, both in the dry and the wet season, and both at the taxon level and the food category level (all with P < 0.01). As shown by the scatterplots (Figure 3), there appears to be no clear relation between body mass and any of the diet quality parameters. Compared with the faeces of the ruminant species, hippo faeces have relatively low nitrogen and phosphorus concentrations and high fibre concentrations. When all species are considered, significant relationships are found by the Spearman rank correlation analyses between body mass and phosphorus (dry season r = −0.41, P < 0.05; wet season r = −0.70, P < 0.01) or nitrogen (wet season only) (r = −0.10, P < 0.05). No significant relation could be found between body mass and fibre concentrations (dry season NDF r = 0.00, P = 0.85; wet season r = 0.60) or nitrogen in the dry season (r = −0.19, P = 0.10). When the analyses focus on ruminants only (thus leaving out hippo), none of the relationships was significant in either season (dry season: NDF r = 0.00, P = 0.85, nitrogen r = 0.21, P = 0.93 and phosphorus r = −0.11, P = 0.66; wet season: NDF r = 0.60, P = 0.17, nitrogen r = 0.80, P = 0.12 and phosphorus r = −0.40, P = 0.33).
Figure 3. Scatterplots showing the relation between body mass (kg) and faecal nitrogen, phosphorus and neutral detergent fibre (NDF) in the dry (a, c, e) and wet (b, d, f) season.
DISCUSSION
Diet selection and resource partitioning
In contrast to the Jarman–Bell principle and recent data on ungulates in a South African savanna (Codron et al. Reference CODRON, LEE-THORP, SPONHEIMER, CODRON, DE RUITER and BRINK2007), our data on West African herbivores do not support the hypothesis that smaller herbivores select forage of higher quality than do larger herbivores. However the results of our study support the species richness–niche width theory of MacArthur & Levins (Reference MACARTHUR and LEVINS1967). They suggested herbivore species richness to be a function of resource availability and niche relationships among species, but they did not address food quality as a factor determining resource partitioning.
Although there appeared to be negative relationships between body mass and nitrogen or phosphorus in the faeces, most relationships were no longer significant or even turned positive when hippopotamus was left out of the data analyses. In both the dry and the wet season, nitrogen and phosphorus concentrations of hippopotamus faeces were much lower than of the other species. In contrast to antelopes and buffalo (family Bovidae), hippopotamus is a non-ruminating foregut fermenter, and its digestive physiology is characterized by relatively long gut retention times and low dry matter digestibility compared with ruminants (Clauss et al. Reference CLAUSS, SCHWARM, ORTMANN, ALBER, FLACH, KHÜNE, HUMMEL, STREICH and HOFER2004, Schwarm et al. Reference SCHWARM, ORTMANN, HOFER, STREICH, FLACH, KHÜNE, HUMMEL, CASTELL, SCHWARM and CLAUSS2006). Bacterial fermentation in the very short hindgut of the hippo is low compared to ruminants (Schwarm et al. Reference SCHWARM, CLAUSS, FLACH and TACK2003, Reference SCHWARM, ORTMANN, HOFER, STREICH, FLACH, KHÜNE, HUMMEL, CASTELL, SCHWARM and CLAUSS2006) and the lower nitrogen concentrations in hippo faeces thus most likely reflect low losses of bacterial protein rather than a general body mass–diet quality relationship.
Although various studies have argued that fibre digestion increases with body mass (Gordon & Illius Reference GORDON and ILLIUS1994, Reference GORDON and ILLIUS1996; Iason & Van Wieren Reference IASON, VAN WIEREN, Olff, Brown and Drent1999, Illius & Gordon Reference ILLIUS and GORDON1992, Robbins et al. Reference ROBBINS, SPALINGER and VAN HOVEN1995), we could find no clear relation between fibre concentrations and body mass. Likewise, Pérez-Barberia et al. (Reference PÉREZ-BARBERIA, ELSTON, GORDON and ILLIUS2004) could find no significant effect of body mass of ruminants on fibre digestibility. In spite of an increase in gut capacity in large species compared to small herbivores, gut retention time may be largely independent of body mass in grazing ruminants (Clauss et al. Reference CLAUSS, SCHWARMM, ORTMANN, STREICH and HUMMEL2007) so that a larger body mass does not necessarily lead to improved digestive efficiency. Rather, the digestive efficiency of fibre is more likely to be related to the proportion of grass versus browse in the diet (Pérez-Barberia et al. Reference PÉREZ-BARBERIA, ELSTON, GORDON and ILLIUS2004).
In line with previous studies (Codron et al. Reference CODRON, LEE-THORP, SPONHEIMER, CODRON, DE RUITER and BRINK2007, Mysterud Reference MYSTERUD2000), there appeared to be no effects of body mass on the food plants that were selected. Similar results were reported by Mysterud (Reference MYSTERUD2000), who found no relationship between the degree of diet overlap and body mass among northern European ungulates, and Codron et al. (Reference CODRON, LEE-THORP, SPONHEIMER, CODRON, DE RUITER and BRINK2007), who found no relation between body mass and the proportion of grass in the diet. It should be noted that in our study diet composition does not merely reflect the proportion of grass versus browse, but refers to (1) the grass genera in the diet, and (2) different food categories such as grass stems, fruits, leaves etc. In addition to the above-mentioned studies, Hansen et al. (Reference HANSEN, MUGAMBI and NAUNI1985) reported no relationship between body mass and dietary diversity of East African ungulates; when re-analysing their data using a Mantel test, no significant correlation can be found between differences in body mass and diet in both the dry (Z = 0.38, P = 0.93) and the wet season (Z = 0.19, P = 0.76).
In the dry season the diets of all herbivores, except the unidentified small species, consisted mainly of grasses, followed by old grass stems and sheaths, and the leaves, flower buds and fruits of Isoberlinia doka. The latter were consumed in particular by zebu, probably because herdsmen are known to regularly offer lopped branches of I. doka trees to their cattle. In the wet season the selection for grasses was much higher, as the diets of all species except zebu cattle consisted of >90% grass. No I. doka leaves were consumed in the wet season, although fruits or flower buds of this tree still made up a considerable proportion of the diet of zebu. Although ripe I. doka pods are woody, young pods contain pulp (Aubréville Reference AUBRÉVILLE1970, Richter & Cumming Reference RICHTER and CUMMING2006) which was found along with pod epidermis. We could not identify any of the seed parts found in the dung. The high proportion of flower buds in the diet of the unidentified small herbivore identifies this species as either common duiker (Sylvicapra grimmia) or red-flanked duiker (Cephalophus rufilatus).
Comparison of the observed dietary overlap with randomized data showed that the overlap in diet between different herbivore species was much higher than what would be expected on the basis of chance. A high degree of dietary overlap can be interpreted in two different, and exactly opposite, ways (Gotelli & Graves Reference GOTELLI and GRAVES1996, Mysterud Reference MYSTERUD2000). First, it can mean that competition is intense because species use the same resources. Alternatively, it may indicate that competition is absent, permitting species to share abundant resources (Gordon & Illius Reference GORDON and ILLIUS1989, Gotelli & Graves Reference GOTELLI and GRAVES1996, Mysterud Reference MYSTERUD2000, Putman Reference PUTMAN1996). The crucial difference is whether or not the shared resources are limiting. Although we did not specifically test this hypothesis, the generally low animal densities, the seemingly abundant supply of grass and the good physical condition (estimated visually) of herbivores even at the end of the dry season showed no apparent indication of limited resources. In addition, herbivore species in Bénoué NP have been shown to differ in patterns of habitat use and spatial distribution in the burned landscape during the dry season (Klop et al. Reference KLOP, VAN GOETHEM and DE IONGH2007, Mayaka Reference MAYAKA2002). In this respect it is interesting that Namgail (Reference NAMGAIL2009) and Namgail et al. (Reference NAMGAIL, FOX and BHATANAGAR2004) demonstrated that blue sheep (Pseudois nayaur) in assemblage with one additional sympatric species, the Ladakh urial (Ovis vignei vignei), showed larger niche width and more niche overlap with abundant food resources, compared with a situation with less abundant food resources and two additional sympatric species, feeding on the same resource.
In our study the niche breadth of food categories consumed by five different herbivore species was much lower in the wet season than in the dry season. This was mainly caused by a higher specialization on grasses in the wet season, and lower proportions of old grass stems and other food categories in the wet season diets. The exception to this pattern is zebu cattle, for which differences in diet composition between the dry and wet season are mainly due to changes in the proportions of leaves and flower buds of Isoberlinia doka in the diet. However, as mentioned above, this may not be behavioural selection by the animals but the result of what is offered by herdsmen. Patterns in niche breadth for grass genera consumed are slightly different than those for food categories. Kob and roan showed major decreases in niche breadth of grass genera consumed during the wet season compared with the dry season, whereas the niche breadth of zebu and in particular hippopotamus greatly increased in the wet season. The patterns shown by kob and roan can be explained by a strong selection for Andropogon species during the wet season, whereas the proportions of different grass species in their dry-season diets were much more balanced. In contrast, in terms of grass genera the wet-season diet of hippopotamus was more balanced than its dry-season diet, and the high proportion of Elymandra grasses was no longer visible in the wet season. As is clear from the resource overlap analyses, at the grass taxon level all species showed much higher overlap in the wet season compared with the dry season. This is probably caused by high proportions of Andropogon in the wet-season diets of all species, which in the dry season is mostly consumed by hartebeest (which has an index of 12.5% for Andropogon, which is between 3.5 to 25 times more than in other species’ diets) but not by the other species.
In practice, actual resource competition between large herbivores is difficult to demonstrate, but the presence of overlap in diet (and niche) in combination with similarities in habitat use and spatial distribution, are expected to alleviate competition. However we conclude that in our study area, due to abundant resources and low herbivore densities, especially in the wet season, niche breadth and diet overlap remain large, because there is no need for herbivores to specialize, to avoid competition over scarce resources.
We suggest that, when herbivore densities in our study area would be higher the hypothesis of Gordon & Illius (Reference GORDON and ILLIUS1996) applies, that large animals may cope with poor-quality diets and thereby avoid competition with smaller species by having a wider resource-tolerance range, especially during limited periods. In our study area, with low herbivore densities and abundant supply of resources, competition seems absent, and all herbivores can optimize diet quality, independent of body size. We conclude that in a situation of abundant resources and low herbivore densities, resource partitioning and niche segregation are ruled by other factors than diet quality and diet composition alone, such as water dependency, presence of predators and human disturbance.
Data considerations
Faecal analyses have been widely used to study herbivore diet composition and quality (Leslie et al. Reference LESLIE, BROWN and JENKS2007). However, as also shown by our study, interspecific differences in food processing or digestive efficiency can lead to major differences in faecal nutrient or fibre concentrations. Several studies have suggested that fibre digestion efficiency increases with the proportion of grass in the diet (Pérez-Barberia et al. Reference PÉREZ-BARBERIA, ELSTON, GORDON and ILLIUS2004), and grazer–browser comparisons based on data from faecal analyses should be made with caution (Codron et al. Reference CODRON, LEE-THORP, SPONHEIMER, CODRON, DE RUITER and BRINK2007). In addition, faecal protein or fibre concentrations can differ significantly between ruminants and non-ruminants due to different digestive strategies (Clauss et al. Reference CLAUSS, SCHWARMM, ORTMANN, STREICH and HUMMEL2007, Schwarm et al. Reference SCHWARM, ORTMANN, HOFER, STREICH, FLACH, KHÜNE, HUMMEL, CASTELL, SCHWARM and CLAUSS2006).
Epidermis patterns are most specific in leaves where they can be identified down to species or genus. There is a clear distinction between the epidermis patterns of grasses and non-grasses (Metcalfe Reference METCALFE1960) but related species cannot always be separated down even to genus as leaf patterns tend to resemble those of related plants. It is not surprising therefore that clusters of look-alikes occur. Usually in those look-alikes the complete leaf pattern is not similar but there are some striking details and as epidermis fragments in herbivore dung were very small (average 0.03 mm2 except 0.09 mm2 for hippo) identification may sometimes be doubtful. This was especially the case for the genera Heteropogon, Hyparrhenia and Schizachyrium, all belonging to the same subfamily (Panicoideae), that all show more or less similar papillae and dumbbell-shaped silica cells. Different vein patterns occur in the genera Aristida, Zea, Chloris, Ctenium, Cynodon, Dactyloctenium, Eleusine, Eragrostis or Sporobolus, most of them from the subfamily Chloridoideae.
Conclusions
Our data suggest that grazing herbivores in Bénoué NP do not segregate in food quality or diet composition along a body mass axis. The lack of a body mass–diet quality relationship is in sharp contrast to the widely supported Jarman–Bell hypothesis (Bell Reference BELL1971, Jarman Reference JARMAN1974). Although several studies have demonstrated an effect of body mass on diet quality of African ungulates, such as recently by Codron et al. (Reference CODRON, LEE-THORP, SPONHEIMER, CODRON, DE RUITER and BRINK2007), the relation between digestive efficiency and body mass may be questionable (Clauss et al. Reference CLAUSS, SCHWARMM, ORTMANN, STREICH and HUMMEL2007, Pérez-Barberia et al. Reference PÉREZ-BARBERIA, ELSTON, GORDON and ILLIUS2004).
As shown here, there can be considerable overlap in the diet composition among mammalian savanna herbivores. The high overlap does not necessarily also imply nutritional overlap since different herbivore species can select post-fire regrowth of different age (Klop et al. Reference KLOP, VAN GOETHEM and DE IONGH2007) and hence, different nutritional quality (Van de Vijver et al. Reference VAN DE VIJVER, POOT and PRINS1999). Although our study provides insight into the importance of body mass, diet quality and diet composition on patterns of resource partitioning among West African herbivores in the dry season and early wet season, further research is needed to investigate patterns of resource partitioning deeper into the wet season, when many herbivores graze on the grazing lawns of hippopotamus (Verweij et al. Reference VERWEIJ, VERRELST, LOTH, HEITKÖNIG and BRUNSTING2006) instead of the post-fire regrowth that is abundant in the dry season.
ACKNOWLEDGEMENTS
This study was supported by the Directoraat Generaal Internationale Samenwerking (DGIS) of the Dutch government and the Institute of Environmental Sciences of Leiden University. The Centre d'Etude de l'Environnement et du Developpement au Cameroun (CEDC), the Ministry of Environment and Forestry (MINEF) and Campement du Buffle Noir provided help and support throughout. In particular we would like to thank Stephen van der Mark and Assan Gomse, and we are grateful to Roeland Koelewijn, Tom Peppink and all park guides who helped collecting the dung samples.
