Within-population isotopic niche variation of mammalian herbivores over a body mass and habitat gradient Authors: Runè van der Merwe, Chanel Lewis, Daryl Codron Published: 2026 ABSTRACT AbstractBackgroundIndividuals within populations may have distinct ecological niches, leading to a partitioned population niche structure. Such structure occurs because of both extrinsic and intrinsic factors. Body mass is suggested to limit between-individual diet niche variation within large-bodied mammalian herbivore populations, as their high intake requirements and associated search costs restricts opportunity for individuals to specialise. This study used stable isotope analysis to investigate niche structure within populations of mammalian herbivores to determine if (1) body mass correlates with the degree of niche structure and if (2) environmental conditions mediate this relationship. Stable carbon and nitrogen isotope compositions of faecal material of 16 herbivore species ranging in body mass (~ 10 to ~ 1420 kg) and geographical location were analysed. For each population, isotopic niche width (ellipse areas) and dispersion patterns (variance in geometric distance between individuals) in isotopic bi-space were quantified as proxies for population niche structure. Linear mixed models were used to test the effect of body mass, and local rainfall and reserve area (proxy for ecological conditions) on these indices of structure.ResultsIsotopic niche widths and dispersion metrics varied between species both within and between reserves, but the directionality of differences was not consistent, with some populations having broader niche widths and/or higher levels of dispersion than others, even within the same species. Although species differed in the degree of niche structuring within their respective populations, there was no effect of interspecific differences in body mass. The degree of dispersion (and thus between individual isotopic niche differences) is also not a function of ecological opportunity, but increased area might allow populations to expand their isotopic niche widths.ConclusionsThese findings suggest that niche variation and dispersion patterns in herbivore populations is limited by factors other than body size and intake requirement, despite the well-established role of body mass in shaping diet, as well as many other behavioural, ecological, and life history traits.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12862-026-02528-6.Keywords: Faeces, Dispersion metrics, Isotopic niche width, Niche structure, Stable isotope analysis Background Niche partitioning is an important process in community assembly, and contributes substantially to species coexistence and the maintenance of biodiversity [1–5]. Partitioning occurs when species narrow their niches to occupy separate components of the resource or other niche axes, thereby reducing overlap and the strength of competition effects [6–8]. The systematic arrangement of species gives communities a distinctive niche structure [9, 10]. Although most commonly studied at community level, a similar partitioning structure pattern can also emerge within populations because individuals may utilise different components of the population’s total resource [11–13]. This individual niche variation means that individuals are non-randomly dispersed within the available niche space, giving niche structure to the population. Actually, the niche variation hypothesis of Van Valen [11] posits that such structure is likely pervasive over time, i.e., individuals become specialist users of a narrow range of resources, a phenomenon that has been demonstrated in a wide range of animal populations [13, 14–16]. Niche diversity among individuals, whether transient or long-term, promotes individual fitness, with significant outcomes for the regulation of populations, and of communities [13, 17–20]. Individual niche variation occurs as a result of both intra- and interspecific competitive interactions that force individuals of the same or different species to use specialised niches [21–23]. Predation can also play a role, both indirectly – by limiting prey populations, predators can alter the level of intraspecific competition downwards, and directly by restricting risk-averse prey individuals to a (narrowed) selection of safer environments [22, 24, 25]. These effects of biotic interactions on structuring are also mitigated by environmental conditions. The “ecological opportunity” hypothesis (referring to the diversity of resources available to consumers) implies that the degree of niche differentiation can only increase in heterogeneous environments that provide opportunity for individuals to diversify in their resource use [22, 26]. Apart from effects of these extrinsic factors, the level of niche variation among individuals within a population can also depend on the intrinsic (life history) traits of species [13, 19, 27, 28]. This was shown amongst thin-toed frogs Leptodactylus spp, in which the degree of individual niche variation of populations differed across two species even within the same environment [29]. Of the various life history characteristics of animals, body mass is widely acknowledged as one of the most biologically important, as it directly regulates behavioural and physiological responses, or acts as a proxy for them because the majority of traits are allometrically linked to it. At the individual level, body mass influences fine-scale processes like habitat, patch, and even bite selection [30–32]. Thus, there is reason to believe that body mass is a characteristic that contributes significantly to within-population niche variation of different species. It has, for instance, been suggested that large-bodied species, like elephants Loxodonta africana, cannot specialise on a limited resource range because the costs of searching for specific foods, while still maintaining sufficient daily intake, are too high [33]. Given that daily intake requirements scale allometrically with body mass [34], we can, by extension, expect that the level of within-population niche structuring varies across species as a decreasing function of body mass. Support for the prediction that intake requirement constraints the potential for individual niche differences comes from the comparative levels of individual niche variation found in carnivorous versus herbivorous mammals [35]. Amongst mammalian carnivores, evidence for extensive individual niche variation within populations has been found in both marine and terrestrial species [14, 15, 36, 37–40]. Amongst mammalian herbivores, which have daily intake requirements about tenfold that of carnivores, only a moderate degree of within-population niche structure was found for the springbok Antidorcas marsupialis and the oryx Oryx gazella [16, 41], and overall such structuring seems to be less prevalent amongst herbivores than in carnivores [35]. Obtaining data appropriate for comparing individual niche positions requires that multiple individuals be sampled commensurately. Stable isotope analysis is a widely-used approach for documenting diet and/or habitat use patterns of animals, and thus well-suited for comparing patterns of niche dispersion amongst individuals within populations [42, 43]. The approach is based on the premise that 13C/12C and 15N/14N ratios measured in the tissues and faeces of organisms reflect that of the dietary resources consumed [44]. In African savanna ecosystems, the foods (plants) of herbivores are bimodally-distributed in stable carbon isotope space, with C3 vegetation (woody plants, forbs, and other dicots) being consistently depleted in 13C relative to C4 photosynthesizing plants grasses and some sedges; [45–47]. Since these two food groups are commensurate with the two major functional food groups that define herbivore evolutionary niche diversification (browse and grass, respectively), stable carbon isotope analysis has already provide key insights into herbivore ecology at all levels of organization, from communities down to individuals [43, 47, 48]. Carbon isotope ratios can also vary due to microhabitat conditions, e.g., plants growing in drier, more exposed sites often show slightly higher δ¹³C values than those in shaded or well-watered areas [49, 50]. Further information about niche separation within (and between) populations can come from stable nitrogen isotope analysis. Nitrogen isotope compositions of mammalian herbivore tissue and faeces are influenced by a variety of factors, including dietary 15N/14N compositions, nutritional properties of the diet such as protein content and digestibility, and physiological stress [43, 51, 52]. For example, herbivores living in arid conditions or during drought stress generally have higher δ15N values due to physiological processes associated with water stress and the excretion of urine with a high 14N enriched urea concentration [51, 53, 54]. Analysed together, carbon and nitrogen isotope data act as a proxy for two interacting trophic niche axes that can be used to identify broad-scale patterns of separation in niche space [55–57]. This is easily visualized – and geometrically quantified – from isotopic biplots, depicting both the isotopic niche width as well as the pattern of dispersion among elements, e.g., among individuals [58, 59]. Here we use stable carbon and nitrogen isotope data from faeces of herbivores, sampled at ten protected areas within four different biomes of southern Africa, to study the degree of within-population isotopic niche structure over a body mass gradient. These data represent 16 species over a body mass range spanning three orders of magnitude, which should be sufficient to test the hypothesis that the level of niche difference within populations is negatively related to body mass. Moreover, we test whether patterns are consistent across the four biomes, or whether the different habitat types mediate the expected relationship. This prediction is grounded in the idea that larger-bodied species are expected to show lower isotopic niche variation at the population level as greater intake requirements necessitate feeding over a broad range of resources. As a result, isotopic values from faecal samples will reflect an averaged diet, producing lower dispersion of isotope values within populations. In contrast, smaller species can meet their dietary requirements in more specialised ways, which may result in greater differentiation between individuals within a population, and thus greater dispersion in isotopic bispace. Methods Study areas and sampling protocol Faecal material was collected from ten different parks and reserves, distributed over four different biomes in southern Africa (Fig. 1; and see descriptions presented in Table 1). Samples were collected between 2013 to 2015 and 2018. Sampling days per site varied between one to four days at each location, depending on the size of the area and the ability to locate animals. Due to logistical constraints we did not resample, i.e., each site was only visited once over the 4-year period of sample collection. Collections of material were mainly in the dry season, with the exception of three sites, because it is expected that niche variation should be more pronounced as there is lower resource availability and consequently increased competition between mammal herbivores. Information on the characteristics, dominant plants and number of large mammals in each of the reserves included in the study is available in Additional file 1, Table S1. Annual rainfall data and detailed information on the time and duration of collection for the ten sites used can be seen in Additional file 1, Table S2. Distribution of the different biomes and locations in southern Africa included in the study Table 1Biome and habitat features of the locations included in the studyBiomeHabitat featuresLocalitySucculent-KarooSemi-desert area with a mild climate that receives mean annual of 168 mm rainfall, mostly during winter. The biome’s most distinguishing feature is its vast diversity of dwarf leaf-succulent shrubs, predominantly Aizoaceae and Crassulaceae.Namaqua National ParkNama-KarooThe semi-arid to arid biome that covers the western interior of the country. This region receives summer rainfall – average annual precipitation of 208 mm - and is characterised by dwarf shrubland, which includes grasses, succulents, geophytes, and annual forbs, but predominantly the Asteraceae, Fabaceae and Poaceae.Augrabies Falls National ParkSavannaStretches to the north and east of South Africa and is the biggest biome in this region. It falls in the summer rainfall zone, with a mean annual rainfall of 495 mm, and the vegetation structure is characterised by a continuous grassy understory and a discontinuous woody canopy. Dominant plants include the Panicoideae, Caesalpinioideae, and Bombacoideae.Musina Nature Reserve; iSimangaliso Wetland Park; Umgeni Valley Nature Reserve; Weenen Game ReserveGrasslandLocated in the elevated interior of South Africa, also receives summer rainfall with a yearly average of 661 mm. The vegetation type is distinguished by a single grass-dominated (Poaceae, predominantly Panicoideae) layer.Willem Pretorius Game Reserve; Sandveld Nature Reserve; Tussen-die-Riviere Nature Reserve; Cradle Nature ReserveInformation from [60] and [61]. Further details in Additional file 1, Table S1 Biome and habitat features of the locations included in the study Information from [60] and [61]. Further details in Additional file 1, Table S1 Because of the limited time spent on each reserve, samples were collected opportunistically. A vehicle was used to locate and observe the study species. Once the animals left the area, faecal samples were collected on foot and placed in brown paper bags. This ensured freshness and no contamination of samples, and the correct attribution to season and species, while also minimizing risk of pseudoreplication. Sample collection was dependant on the identity of the species. For herd-forming species like wildebeest Connochaetes spp, care was taken to avoid overrepresentation of individuals from a single herd by distributing sampling effort across different groups. In smaller reserves, where few distinct herds were present, such distribution was not always feasible, contributing to variation in samples sizes. For species that are more difficult to locate, such as kudu Tragelaphus strepsiceros, all available samples were collected when individuals were encountered, further contributing to uneven samples sizes across species. In smaller reserves sampling was completed within a single day, while in larger reserves, sampling spanned multiple days and areas, ensuring spatial distribution of individual samples. Faecal material used in this study (n = 460) included 16 different species of free-ranging herbivores, comprising three feeding guilds – browsers, grazers, and intermediate-feeders – and ranged over a body mass gradient of ~ 10 kg to ~ 1 420 kg (Table 2). See Additional file 1, Table S3 for detailed sample sizes per reserve. Table 2List of species used for isotopic analysis of faecal materialScientific NameCommon Name n Body MassDietDigestion%Grass Raphicerus campestris Steenbok710.5BRRuminant10.0 Antidorcas marsupialis Springbok5139.1IMRuminant30.0 Aepyceros melampus Impala2052.5IMRuminant60.0 Phacochoerus africanus Common warthog2082.5GRHindgut95.0 Tragelaphus angasii Nyala886.6IMRuminant20.0 Damaliscus pygargus Blesbok8102.0GRRuminant100.0 Oryx gazella Gemsbok6169.0GRRuminant75.0 Alcelaphus buselaphus Red hartebeest55171.0GRRuminant97.0 Connochaetes gnou Black wildebeest20180.0GRRuminant81.0 Connochaetes taurinus Blue wildebeest33180.0GRRuminant87.5 Kobus ellipsiprymnus Waterbuck28210.0GRRuminant84.0 Tragelaphus strepsiceros Kudu43213.5BRRuminant5.0 Equus quagga Plains zebra74276.0GRHindgut90.0 Tragelaphus oryx Eland20570.0BRRuminant10.0 Giraffa camelopardalis Giraffe61900.0BRRuminant0.9 Hippopotamus amphibius Hippopotamus181417.5GRForegut95.0n is the number of individual specimens collected for this study; body mass is the species’ mean given in kilograms (references from [62]); Species were classified according to foraging guild as browsers (BR), grazers (GR) and species with an intermediate diet (IM); Digestion; %Grass is the mean percent of diet expected to consist of grass (references from [63]) List of species used for isotopic analysis of faecal material n is the number of individual specimens collected for this study; body mass is the species’ mean given in kilograms (references from [62]); Species were classified according to foraging guild as browsers (BR), grazers (GR) and species with an intermediate diet (IM); Digestion; %Grass is the mean percent of diet expected to consist of grass (references from [63]) Laboratory preparations The samples were air-dried, and a coffee grinder was used to homogenise dried faecal samples to a fine powder. Thereafter, between 1.1 and 1.2 mg of each specimen was weighed into small tin capsules. Specimens were combusted in an elemental analyzer (Flash EA 1112 Series) to release N2 and CO2 gas. An automated ConFlo IV inlet system, attached to a Delta V Plus stable light isotope ratio mass spectrometer (all equipment supplied by Thermo Fischer, Bremen, Germany) was used to measure 13C/12C, and 15N/14N. Results are reported in delta (δ) notation in parts per mil (‰), relative to the Vienna PeeDee Belemnite (VPDB) and atmospheric N2 standards for carbon and nitrogen isotopes, respectively. The standard deviation of replicates of the internal reference materials (Merck Gel and DL-Valine) was less than 0.1‰ for both δ13C and δ15N. Statistical analysis All statistical analyses were conducted with the statistical software R version 4.3.2 [64]. For analysis, we defined each species per reserve as a distinct population. The sample size (n) representing each population ranged from 1 to 20; those represented by small samples sizes (n < 6) were excluded from further analysis. To estimate isotopic niche width of each population, we used Bayesian estimates of the Standard Ellipse Area in isotopic bi-space (SEAb), computed using SIBER, based on the means estimated from 100 000 iterations, with a burn-in of 1 000, and thinning of 10 [58]. The SEAb (measured in units ‰2) accommodates approximately 40% of the δ13C and δ15N data, and so measures the core nucleus of the isotopic niche of a population. However, while SEAb is a widely used and robust metric, it can underestimate niche variation in populations where individual diets are highly generalised or overlapping, resulting in small ellipse areas. To compliment this, we quantified the degree of dispersion amongst individuals in isotopic bispace, as a proxy for within-population niche structure. The following metrics used: (1) meandistance to the centroid (CD); (2) mean nearest neighbour distance (NND); and (3) standard deviation of the nearest neighbour distance (SDNND); [58, 59], also calculated using the SIBER [59] package. These metrics are commonly used indicators of properties of the isotopic niche structure of whole communities [59]. However, Euclidean geometry such as nearest-neighbour distances have long been used in spatial ecology to describe dispersion patterns within populations as well [65]: the largest NND occurs in uniformly dispersed populations as competitive interactions keep individuals further apart in geometric space, whereas small NND is associated with individual aggregations. In terms of dispersion in isotopic bi-space, large CD and NND can thus be associated with a high degree of individual separation and thus within-population niche structure, and low values indicate greater niche similarity amongst individuals. Larger species are expected to have fewer individual niche differences, and so less defined within-population niche structure than smaller species. Thus, we hypothesized that the pattern of dispersion in isotopic bi-space would vary from aggregated (low CD, NND, and SDNND) to uniform (high CD, NND, and SDNND) across species of decreasing body size. Consequently, a negative relationship should arise between body mass (BM) of large herbivores and the metrics CD, NND, SDNND, as well as SEAb. We initially evaluated these relationships using Pearson correlations. However, to account for the multi-levelled data structure, i.e., the presence of multiple sampling localities and biomes (see Table 2), these relationships were also evaluated using linear mixed models (LMMs), including ‘Site’ and ‘Biome’, the former nested within the latter, as random effects. The dependent variables ‘CD’, ‘NND’, ‘SDNND’ and mean ‘SEAb’, as well as the predictor variable ‘BM’, were all ln-transformed to achieve normality and homoscedasticity of residuals. Various model combinations were fit to the data and compared using AIC (see below), i.e., models including and excluding either or both random effects, including or excluding ‘BM’, and also those including random slopes (interactions between BM and the random effects). In some models, the variable BM was replaced with a qualitative predictor ‘Species’ to test for differences between species that might arise due to reasons other than variations in BM. To address habitat effects, i.e., the ecological opportunity hypothesis, more directly, LMMs were used where the qualitative habitat variables ‘Biome’ and ‘Site’ were replaced with quantitative predictors linked to productivity, specifically the area (in hectare; ha) of each reserve, and climate data i.e., mean annual rainfall and minimum and maximum temperatures, all of which act as proxies for spatial variation in resource supply, i.e., vegetation biomass [66]. Pearson’s correlation was conducted to test for multicollinearity between rainfall and temperature. These variables were correlated (P < 0.01, r = -0.7167), and therefore only rainfall was included in the models. Models included various combinations of ‘Rainfall’, ‘Area’, and their interaction, with ‘Species’ as a random effect for comparison. Again, the dependent variables, as well as ‘Area’ and ‘Rainfall’, were ln-transformed to satisfy parametric test assumptions. For both sets of models, null models (i.e., an intercept-only model) were included for comparison, and the percentage variance explained by each variable was computed using Variance Component Analysis (VCA). Models were compared based on the Akaike’s Information Criterion (AIC; [67]) adjusted for small samples (AICc). For each model ΔAICc was calculated according to: The model with a ΔAICc of 0 is considered to fit the data best. To separate competing models, we followed guidelines from Burnham and Anderson [68], where models with a ΔAICc < 2 have substantial support, values between four and seven suggest considerably less support and for ΔAICc >10 models are considered to have essentially no support. All LMMs were fit with the ImerTest package [69]. Results Four hundred and sixty faecal samples from 16 herbivore species, collected from ten reserves distributed across four South African biomes, were analysed for stable carbon and nitrogen isotope compositions. The body mass of species ranged from ~ 10 kg to ~ 1420 kg and included populations sampled over a habitat gradient, from more xeric (receiving ~ 188 mm of rain annually) to mesic (receiving ~ 1030 mm annually) environments. A representation of isotopic δ13C and δ15N data for the herbivore populations sampled can be seen in Fig. 2, also illustrating the isotopic niche areas (SEAb; ‰2) of populations. There are clearly differences in both carbon and nitrogen isotope compositions between species within reserves, as well as between populations of the same species across reserves. Stable Carbon and Nitrogen Isotope biplots illustrating the isotopic distribution of mammalian herbivores across (a) Namaqua National Park, (b) Augrabies Falls National Park, (c) Tussen-Die-Riviere Nature Reserve, (d) Willem Pretorius Game Reserve, (e) Sandveld Nature Reserve, (f) Cradle Nature Reserve, (g) Musina Nature Reserve, (h) isiMangaliso Wetland Park, (i) Weenen Game Reserve and (j) Umgeni Valley Nature Reserve. Ellipses incorporate 40% of the data in each population sampled. Diet indicated by colours and symbols: Green-blue and triangles for grazers; red and circles for browsers; and yellow and squares for intermediate feeders The expected separation of herbivores into different feeding guilds was observed at most sites based on distributions of δ13C (Fig. 2; and see Additional file 1, Table S3 for descriptive statistics and sample sizes per reserve). Grazer δ13C was generally consistent with C4 feeding (mean δ13C of -17.8‰; n = 255), browser δ13C with C3 feeding (mean δ13C of -26.9‰; n = 126), and although it was expected that δ13C of intermediate feeders would plot out in between grazers and browsers, most of them consumed more C3 browse than C4 grasses (Fig. 2; mean δ13C of -25.9‰; n = 79). Mean δ15N was highest for red hartebeest (12.0‰) and springbok (11.8‰) populations from Namaqua National Park, the most arid of all the sites included in the study. Mean δ15N was also generally high for species from Augrabies Falls National Park (ranging between 6.5‰ and 7.8‰), Musina Nature Reserve (between 5.0‰ and 9.7‰), and Sandveld Nature Reserve (extending from 6.3‰ to 7.3‰). Within-population dispersion and isotopic niche widths All three of the dispersion metrics – CD, NND, and SDNND – as well as isotopic niche widths (SEAb), varied widely across species and reserves. Interestingly, the directionality of differences across species was not consistent, and so, contrary to our main prediction, dispersion patterns showed no evident relationship to body mass (Fig. 3; and see below). However, some consistencies in spatial patterns are noticeable: populations from the Grassland biome generally had higher SEAb compared to populations from the other biomes, and populations from the Succulent-Karoo had the lowest SEAb across all biomes. Relationship between isotopic niche metrics and body mass of large mammalian herbivores in four different biomes of southern Africa: Yellow – Succulent Karoo (n = 2); Red – Nama-Karoo (n = 4); Green – Grassland (n = 11) and Orange – Savanna (n = 17); Black – All samples included in the study, irrespective of biome (n = 34), for (a) Distance to the centroid (CD; P = 0.4291, r2 = 0.1402), (b) Nearest neighbour distance (NND; P = 0.9717, r2 = 0.0063), (c) Standard deviation of the nearest neighbour distance (SDNND; P = 0.5785, r2 = -0.0987), and (d) Standard Ellipse Area (SEAb; P = 0.6310, r2 = 0.0854). Isotopic niche metrics and body mass are ln-transformed Body mass and habitat effects Linear mixed models (LMMs) were used to examine the effect of body mass, species, site and biome on the three dispersion metrics and on SEAb (Table 3). For all dispersion metrics, and SEAb, the best-supported models were null models, implying only small or even negligible effects of species identity, biome, and habitat. For CD, the model including body mass only also had substantial support (∆AICc = 1.7), while all other models (for all dependent variables) are only weakly supported (∆AICc > 2). Moreover, regardless of the combination of predictors included in any model, none of the predictors ‘Body mass’, ‘Species’, ‘Site’, as well as the ‘Biome’ from where samples were derived were statistically significant at α = 0.05, and all failed to explain much variation in any of the response variables (marginal R2 was generally smaller than 0.034). Thus, the LMMs confirmed inferences based on visual inspection of Fig. 3 that body mass had no significant effect on any of the dispersion metrics, even after accounting for geographical and taxonomic effects. Based on VCA results, ‘Site’ explained more of the variation than ‘Biome’ and ‘Species’, but neither of these three random terms explained more variance than random error (residuals); at best, ‘Site’ explained only ~ 25% of total variance in NND and SDNND, ‘Biome’ around 6% of variance in SEAb, and ‘Species’ around 9% in NND. Table 3Linear mixed effect models (LMMs) to test the effect of body mass (fixed effect), species, site, and biome (random effects) on dispersion metricsModel∆AICc n KFixed effectRandom effectsBody MassBiomeSpeciesSite CD Null model0342----BM1.73430.0928 ± 0.1158---Biome4.9343-0.0042 (1%)--Species4.9343--0.00 (0%)-Biome/Site6.8344-0.00 (0%)-0.0624 (13%)Species + Biome7.5344-0.0042 (1%)0.00 (0%)-BM+ Biome9.33440.0939 ± 0.11540.0060 (1%)--Species + Biome/Site9.6345-0.00 (0%)0.00 (0%)0.0624 (13%)BM + Biome/Site11.13450.1147 ± 0.11170.00 (0%)-0.0785 (16%) NND Null model0342----BM2.43430.0036 ± 0.1002---Species5.1343--0.0197 (6%)-Biome5.2343-0.00 (0%)--Biome/Site7.5344-0.00 (0%)-0.086 (23%)Species + Biome7.7344-0.00 (0%)0.0197 (6%)-Species + Biome/Site10.1345-0.00 (0%)0.0335 (9%)0.0949 (25%)BM+ Biome10.53440.0036 ± 0.10020.00 (0%)--BM + Biome/Site13.1345-0.0063 ± 0.09730.00 (0%)-0.0764 (20%) SDNND Null model0342----BM2.1343-0.0872 ± 0.1553---Species4.3343--0.0081 (1%)-Biome4.3343-0.00 (0%)--Biome/Site6.7344-0.00 (0%)-0.1720 (20%)Species + Biome6.9344-0.00 (0%)0.0081 (1%)-BM+ Biome8.5344-0.0872 ± 0.15530.00 (0%)--Species + Biome/Site9.4345-0.00 (0%)0.0553 (6%)0.2297 (25%)BM + Biome/Site11.2345-0.0728 ± 0.15340.00 (0%)-0.0878 (10%) SEA b Null model0342----BM2.23430.1011 ± 0.2084---Biome3.3343-0.0856 (6%)--Species3.7343--0.00 (0%)-Biome/Site5.5344-0.00 (0%)-0.1821 (12%)Species + Biome5.9344-0.0856 (6%)0.00 (0%)-BM+ Biome6.93440.1108 ± 0.20460.0896 (6%)--Species + Biome/Site8.2345-0.00 (0%)0.00 (0%)0.1821 (12%)BM + Biome/Site9.23450.1252 ± 0.20230.00 (0%)-0.1958 (13%)n is the number of observations, K is the number of parameters, for the fixed effect BM (Body mass; ln transformed) the parameter esitmate is given with the standard error in brackets. For the random effects variance is given with percentage variance explained by the variable in brackets Linear mixed effect models (LMMs) to test the effect of body mass (fixed effect), species, site, and biome (random effects) on dispersion metrics n is the number of observations, K is the number of parameters, for the fixed effect BM (Body mass; ln transformed) the parameter esitmate is given with the standard error in brackets. For the random effects variance is given with percentage variance explained by the variable in brackets Similar models were also used to test the effects of local rainfall regimes and reserve area (proxies for primary productivity) on the three dispersion metrics and SEAb (Table 4). Here body mass was not included in the LMMs as it did not have any significant effect in the previous set of models. For CD and SDNND, the null models were the best supported. For NND the best fit model included rainfall and species, although not statistically significant at α = 0.05. While area was a significant predictor for SEAb, model selection provided the strongest support for the null model. Table 4Linear mixed effect Models (LMMs) to test for the effect of ecological diversity on dispersion metrics at an annual scaleModel∆AICc n KFixed effectsRandom effectAreaRainfallSpecies CD Null model0342---Rain + Species8.9344-0.0091 ± 0.19830.00 (0%)Area + Species8.93440.1032 ± 0.0746-0.00 (0%)Area + Rain + Species13.13450.1052 ± 0.0767-0.0328 ± 0.19790.00 (0%)Area + Rain + Area*Rain + Species17.3346-0.6887 ± 0.7052-1.2749 ± 1.11440.00 (0%) NND Rain + Species0344--0.0240 ± 0.16810.0223 (6%)Area + Species0.33440.0848 ± 0.0644-0.0489 (14%)Null model1.1342---Area + Rain + Species4.63450.0937 ± 0.0660-0.0943 ± 0.16440.0668 (19%)Area + Rain + Area*Rain + Species6.9346-1.0193 ± 0.5822-1.8086 ± 0.90880.0554 (17%) SDNND Null model0342---Rain + Species27.1344--0.1362 ± 0.26210.0302 (4%)Area + Species28.83440.0671 ± 0.1021-0.0336 (4%)Area + Rain + Species31.93450.0823 ± 0.1045-0.1839 ± 0.26520.0702 (8%)Area + Rain + Area*Rain + Species35.9346-0.8541 ± 0.9633-1.626 ± 1.51540.0439 (5%) SEA b Null model0342---Area + Species43.6344 0.2744 ± 0.1284 -0.00 (0%)Rain + Species45.9344-0.1106 ± 0.35400.00 (0%)Area + Rain + Species46.7345 0.2743 ± 0.1321 0.0014 ± 0.34110.00 (0%)Area + Rain + Area*Rain + Species50.6346-0.6664 ± 1.2287-1.4704 ± 1.94170.00 (0%)n is the number of observations, K is the number of parameters in the model, for the fixed effects Area and Rainfall (ln transformed) the parameter estimates is given with the standard error in brackets, for Species variance is given with percentage variance explained by variable in brackets, and significance is indicated in bold (p < 0.05) Linear mixed effect Models (LMMs) to test for the effect of ecological diversity on dispersion metrics at an annual scale n is the number of observations, K is the number of parameters in the model, for the fixed effects Area and Rainfall (ln transformed) the parameter estimates is given with the standard error in brackets, for Species variance is given with percentage variance explained by variable in brackets, and significance is indicated in bold (p < 0.05) Discussion The degree of individual diet variation and overall niche structure within populations are important factors for our understanding of the mechanisms that drive niche dynamics, competitive interactions, and community assembly, and may be linked to individual fitness, selection processes and ultimately contribute to the maintenance of biodiversity [2, 4, 70–72]. Yet, the driving factors of niche variability within populations are still not clearly understood [22]. That organisms’ body mass play a key role in determining diet, as well as a host of other life history and ecological characteristics, is well known [73–75], and the potential for body mass to regulate individual niche variation within populations has been suggested based on theoretical models and (limited) empirical data [33, 35]. Here, patterns of within-population niche variation across differently-sized mammalian herbivores were investigated using stable isotope analysis of faeces. Despite significant variations between populations and species, including the well-known distinction between C3 browsing and C4 grazing along the δ13C axis, we found no discernible effect of body mass on any measures of within-population niche dispersion, or niche width. The effect of body mass The distribution of individual datapoints in isotopic bi-space, along with geometric indices of isotopic niche dispersion and variability, provides a useful means for investigating niche structure within populations. Given their transient nature, faecal isotopic data do not measure the persistence of individual niche variation (sensu [22]), but do offer an indication of the degree of behavioural dissimilarity across individuals at any one point in time (see below). While a wide range of population niche structures were recovered, species’ position in terms of niche width and degree of dispersion vary stochastically, and body mass (and by extension all life history characteristics closely related to it) did not emerge as a strong predictor of structure, at least among southern African herbivores. Patterns of isotopic niche metrics were not consistent for all populations and species. This inconsistency is one reason why no relationship with body mass emerged. The lack of body mass effect was counter to predictions, but this result does not mean that life history is not important. For one, our initial hypothesis was based on variation in absolute intake requirements only, but there are other life history and behavioural traits that function at the individual level that could further influence diet niches. For example, individuals differ in their needs for nutrients and in their ability to cope with toxins, and in order to meet their required nutrient uptake levels and to avoid overdosing on certain toxins herbivores need to consume a diversity of plant species [76–78]. Therefore, factors like nutrient requirements or biochemical interactions between plants and herbivores might better explain species, and even population-level, differences in herbivore niche variation and structure. Habitat effects on isotopic niche width Variation in isotopic niche metrics across populations may in part reflect differences in the availability of C3 versus C4 plants among biomes. For example, savanna reserves include a mixture of grasses and woody C3 plants, potentially providing greater isotopic variability, whereas the grasslands are dominated by C4 grasses. Populations from the Grassland biome generally exhibited higher SEAb than those from other biomes, which may appear counter-intuitive given the lack of woody vegetation in these landscapes, but a greater degree of mixed C3/C4 feeding has previously been recorded for herbivores in South African grasslands compared to savannas [79]. Thus, it appears that grassland herbivores, even grazing taxa, consume significant amounts of forbs and other ground-level dicots more frequently than their savanna counterparts. However, there was no consistent pattern across all metrics or species, suggesting that variation in C3/C4 plant availability alone does not fully explain differences in isotopic niche structure. Although not ubiquitous, the results shown here offer some evidence for a positive relationship between area of reserves and isotopic niche width. MacArthur’s [80] habitat heterogeneity hypothesis posits that increased habitat diversity within an area creates a greater number of niches, which in turn can support higher species diversity. In the same way, the niche diversity of larger areas likely allows individuals to exploit different resources, reducing direct competition, such that larger, more heterogeneous areas support populations with broader diet niche widths. However, although isotopic niche width (SEAb) broadened, there was no effect of reserve area on any of the dispersion metrics (CD, NND, or SDNND), thus any increase in niche breadth seems to have occurred at population not individual levels. Furthermore, there was no effect of reserve area or rainfall on dispersion metrics, suggesting that the degree of dispersion (and thus isotopic niche differences) is not a function of ecological opportunity. To upscale individual niche variation to the community level in herbivores, we need to understand how population niche breadths expand while dispersion patterns remain unchanged as area (opportunity, i.e., access to wider resource diversity) increases. Answering this question requires extensive longitudinal records of within-individual dietary histories. Limitations The use of faecal matter for stable isotope analysis enables non-invasive sampling and is less time-consuming than traditional approaches. However, faecal matter only provides short-term dietary information [35, 81]. Thus, while the data do represent variation in within-population niche structures, the patterns are obviously transient: they do not record the within-individual component of niche variation (sensu [17] and others). The use of long-term profiles such as teeth, hair or bone is more indicative of persistent dietary niche variation over time [55]. Stable isotope analysis is useful for classifying diet niches at broad scales, such as C3/C4 feeding. But, quantitative analysis as used here may under-represent true dietary niche widths, biasing them downwards when mixed diets occur [82]. While not a major drawback for studies of large herbivores, especially in African savannas where C3/C4 distinctions are consistently related to the diet niche [83], qualitative analyses such as those based on isotope mixing models could add additional insights [84]. Of course, this requires knowledge of local-scale variations in isotopic compositions of a variety of plant foods, as well as diet-consumer fractionation effects that may be species-, diet-, or even regionally-specific. Nonetheless, since we focused on dispersion within populations here, isotopic ‘averaging’ of diets should not bias patterns of individualisation, as inferred from measures like CD and NND. One reason we found no effect of body mass could be that there are too few species within each size class in our data. Data available for African elephants reveal very little isotopic niche variation between individuals [33], whereas data available for rodents reveal high levels of individual specialisation [85]. Species included for the present study exist between these size extremes, but it may be that body mass effects on niche structure only manifest at very large and very small body mass. The inclusion of larger herbivores i.e., buffalo Syncerus caffer, rhinoceroses Ceratotherium simum and Diceros bicornis, and elephants, and smaller herbivores e.g., klipspringer Oreotragus oreotragus and duiker Sylvicapra grimmia, as well as other small herbivores like hyraxes, may be useful to expand the current data and thereby investigate the possibility of polytony in the body mass effect on niche variation. Conclusion Many studies investigating within-population niche variation compares only one or two species, and thus comparative approaches for testing hypotheses about life history effects are scarce. A novel approach was used to study the potential effects of a fundamental life history trait – body mass – on within population niche variation across a range of mammalian herbivores species. The lack of a body mass effect was unexpected as the significance of body mass in diet determination, especially amongst ruminants, is well known. Insights about niche variation and structure within herbivore populations scale upwards to community levels to inform our understanding about niche partitioning and coexistence in these species. To get there we need to understand why population niche breadths expand while dispersion patterns remain unchanged, necessitating longitudinal records of within-individual diet histories. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Acknowledgements We thank Dr Grant Hall (Mammal Research Institute, University of Pretoria) for isotope analysis and the South African Weather Services (SAWS) for climate data provided. We thank two anonymous reviewers for their constructive feedback that helped to improve the quality of this manuscript. Author contributions DC conceived the study and collected field materials. RvdM and CL prepared samples for isotope analysis; RvdM analysed the data. RvdM wrote the first draft of the manuscript. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Funding Funding was provided by the Palaeontological Scientific Trust (PAST; RvdM), the Centre of Excellence in Palaeoscience (Genus; RvdM), and the National Research Foundation of South Africa (Grant UIDs 129172 and 137968; DC). Data availability The isotopic dataset used in this study is available in Figshare with the identifier https://doi.org/10.6084/m9.figshare.20591313. Declarations Ethical approval This study involved no experimental or other manipulation of animals. Sampling included collection of field-discovered faecal matter only, for which ethical norms and standards according to South African guidelines were adhered to (University of the Free State Animal Ethics Research Committee permit number 08/2014/21). Permission for fieldwork and sample collection was also approved by relevant conservation authorities: Free State Department of Economics, Small Business Development, Tourism and Environmental Affairs (FSDESTA; permit number 01/20893), Ezemvelo Wildlife (OP 1698/204), and iSimangaliso Wetland Park. Consent for publication Not applicable. Peer review The Author comments and Reviewer reports are available in supplementary file 2. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement The isotopic dataset used in this study is available in Figshare with the identifier https://doi.org/10.6084/m9.figshare.20591313.