Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Application of Nitrogen and Carbon Stable Isotopes (δ15N and δ13C) to Quantify Food Chain Length and Trophic Structure

Abstract

Increasingly, stable isotope ratios of nitrogen (δ15N) and carbon (δ13C) are used to quantify trophic structure, though relatively few studies have tested accuracy of isotopic structural measures. For laboratory-raised and wild-collected plant-invertebrate food chains spanning four trophic levels we estimated nitrogen range (NR) using δ15N, and carbon range (CR) using δ13C, which are used to quantify food chain length and breadth of trophic resources respectively. Across a range of known food chain lengths we examined how NR and CR changed within and between food chains. Our isotopic estimates of structure are robust because they were calculated using resampling procedures that propagate variance in sample means through to quantified uncertainty in final estimates. To identify origins of uncertainty in estimates of NR and CR, we additionally examined variation in discrimination (which is change in δ15N or δ13C from source to consumer) between trophic levels and among food chains. δ15N discrimination showed significant enrichment, while variation in enrichment was species and system specific, ranged broadly (1.4‰ to 3.3‰), and importantly, propagated variation to subsequent estimates of NR. However, NR proved robust to such variation and distinguished food chain length well, though some overlap between longer food chains infers a need for awareness of such limitations. δ13C discrimination was inconsistent; generally no change or small significant enrichment was observed. Consequently, estimates of CR changed little with increasing food chain length, showing the potential utility of δ13C as a tracer of energy pathways. This study serves as a robust test of isotopic quantification of food chain structure, and given global estimates of aquatic food chains approximate four trophic levels while many food chains include invertebrates, our use of four trophic level plant-invertebrate food chains makes our findings relevant for a majority of ecological systems.

Citation: Perkins MJ, McDonald RA, van Veen FJF, Kelly SD, Rees G, Bearhop S (2014) Application of Nitrogen and Carbon Stable Isotopes (δ15N and δ13C) to Quantify Food Chain Length and Trophic Structure. PLoS ONE 9(3): e93281. https://doi.org/10.1371/journal.pone.0093281

Editor: Petra Quillfeldt, Behavioural Ecology & Ecophysiology group, Denmark

Received: September 1, 2013; Accepted: March 4, 2014; Published: March 27, 2014

Copyright: © 2014 Perkins et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This research was funded by The Food and Environment Research Agency, United Kingdom, under their Seedcorn Programme. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Understanding food web structure is of critical importance to a broad suite of ecological theory given that trophic dynamics between individuals, populations, species and functional guilds underpin the ecological functioning and evolution of biological communities [1], [2]. Quantifying food web structure (trophic structure hereafter) is therefore a prerequisite to better understand how it in turn interacts with emergent properties of organisms and the environment, such as energy flux [3], population dynamics [4], patterns of biodiversity [5] and ecosystem functioning [6], [7]. Determination of feeding relationships between species is integral to quantifying trophic structure, and traditional methods include gut-content analysis, faecal analysis and behavioural observations. However, these methods can be laborious and may not reflect variation in digestibility and assimilation of source items, and if limited in their collection in space and time, may lead to over or under representation of source contributions [8]. Increasingly, stable isotope ratios of nitrogen (N15:N14, termed δ15N) and carbon (C13:C12, termed δ13C) in consumer tissues are used to provide a temporally and spatially integrated construct of dietary niche [8], with δ15N and δ13C of consumer proteins reflecting the proteins of their food sources [9], [10]. Typically, enrichment in δ15N of 2.5‰ to 3.4‰ is observed from diet to consumer [11]–[13], allowing determination of an organism's trophic level [11], [14], [15] and overall food chain length [16], [17]. Conversely, enrichment in δ13C is much smaller between diet and consumer [11], [13], and because basal sources often differ in their δ13C values, δ13C can be used to trace prey–consumer connections or food chains [11]. Hence change in δ15N and δ13C from source to consumer as described (termed trophic discrimination factors and represented as Δδ15N or Δδ13C), is the mechanism that crucially underpins the positioning of individuals, populations and species relative to one another in bivariate isotopic space (typically with δ15N on a y-axis and δ13C on an x–axis). Importantly this subsequently allows for measures of Euclidean distances across the isotopic space occupied by populations, species or communities in order to quantify aspects of trophic structure [18]–[21]. For instance, food chain length is calculated as λ + (nitrogen range/average Δδ15N), where λ is minimum trophic position and nitrogen range (NR) is mean difference between trophic levels of maximum and minimum δ15N [11], [14], [18]. Similarly, carbon range (CR) measures breadth of trophic sources and is calculated using mean difference between maximum and minimum δ13C of community members [18]. Observational studies have largely used such measures to quantify food chain length, typically in aquatic systems, in response to factors such as ecosystem size, disturbance and productivity [22]–[24]. Thus the use of stable isotope ratios in an organism's tissues to provide temporally and spatially integrated dietary data is proving a very valuable methodology for trophic research.

Critically though, variation in Δδ15N and Δδ13C generates uncertainty/error in subsequent estimates of trophic structure and relationships [11], [17], [25]–[28]. Such variation in discrimination factors is well documented in the literature (−0.8‰ to 5.9‰ for δ15N and -2.7% to 3.4‰ for δ13C, excluding fluid feeders; [29]). This can be a consequence of multiple factors, including dietary protein quality, metabolic process and efficiency of protein assimilation and loss, fasting, growth rate, age, size, tissue type, sample size and sampling process, although there is considerable debate on which are most important [12], [13], [27], [30] (and references therein for all). It is the cumulative effect of all such underlying sources of variation that we observe as variation in discrimination factors between different individuals and species. Common practice often uses mean estimates of Δδ15N and Δδ13C ignoring variability around estimates; consequently, derived estimates of trophic structure and subsequent ecological conclusions may lack accuracy.

These issues associated with variance have driven recent innovations in the analysis of isotopic data that provide practitioners with tools to apply Bayesian inference to the calculation of trophic structure metrics [20]. These approaches are ideally suited to testing effects of variance as they provide population and community trophic metric estimates based on resampling of variance in mean δ15N and δ13C estimates, effectively quantifying and propagating variation in raw data as uncertainty in subsequent metric outputs, allowing for critical examination of precision in estimates of trophic structure.

Despite variation in discrimination being widely documented (e.g. [11], [13], [29]) to date relatively few studies have validated isotopic structural measures such as NR and CR against known measures of food chain length, despite repeated calls for research [21], [27]. Of such studies, variation in baseline δ15N has been shown to affect food chain length estimates [17], and variation in Δδ15N to affect trophic level/food chain estimates [11], [25], [28], all in aquatic systems. Such studies have therefore been vital in guiding our understanding and use of these techniques; however, most have used wild systems which may not accurately identify basal sources or preclude omnivory, and thus do not explicitly know/control trophic levels against which to compare isotopic measures of structure. To further improve our understanding of isotopic approaches to quantifying trophic structure, so it would be prudent to additionally test the accuracy of isotopic measures of trophic structure for multi-trophic level food chains when trophic levels are explicitly controlled and food chain lengths known, in addition to allowing for the propagation of variation in Δδ15N and Δδ13C to final isotopic trophic measures.

It is of further importance to understand effects of variation in Δδ15N and Δδ13C upon structural measures such as NR and CR given these univariate measures underpin other bivariate (δ15N with δ13C) measures of trophic structure in isotopic space. NR is also the most used isotopic metric in observational studies [21], and its function as a tool to quantify trophic level, food chain length or as a component of bivariate measures is dependent upon an assumed constant δ15N enrichment with each consumer level. Variance around this assumed average enrichment constitutes unknown error in estimates of NR in observational studies. Thus, experimental validation would improve understanding of the importance of variance in Δδ15N and Δδ13C for affecting quantifications of trophic structure in wild systems [21], strengthening subsequently derived ecological conclusions, in addition to further catalysing the development and use of these techniques to/by a wider audience of ecologists.

In this study, we use natural plant and insect food chains raised under controlled conditions to examine the dynamics of δ15N and δ13C over four trophic levels, and using resampling procedures to allow propagation of variance in Δδ15N and Δδ13C, test the accuracy of the isotopic metrics NR and CR to quantify trophic structure. We use a terrestrial plant-insect system in this study simply because terrestrial systems are currently less well studied relative to aquatic systems and it would therefore broaden our knowledge of isotopic techniques to have a greater range of validation studies to draw upon. By using four trophic levels, we broaden cross–system applicability of our results to a larger repository of aquatic studies, given that global aquatic food chain lengths have been estimated at 3.5 to 4.0 trophic levels [17]. Specifically, we test three questions: 1) How consistent are Δδ15N and Δδ13C between trophic levels and among food chains? 2) Does NR accurately determine food chain length? 3) How does CR change with food chain length?

Methods

Ethics Statement

All wild animals used in this experiment are not protected by law, and were collected on land belonging to the University of Exeter for which permissions to collect had been granted.

Experimental Set-up

To test dynamics of δ15N and δ13C with changing trophic level, three replicate food chains were raised in the laboratory, with a further analogous food chain collected from the wild to allow comparison with wild systems. All food chains had four trophic levels consisting of: primary producer (plant) → herbivore (aphids feeding on plant phloem sap) → predator (hoverfly larvae feeding on aphids) → secondary predator (parasitoid hymenoptera which are obligate endoparasites of hoverfly pupa, with a single parasitoid emerging from a single pupa). These organisms were used as they provided us the opportunity to study natural food chains (that can be found in nature) and were easy to culture in controlled laboratory conditions so that trophic levels were explicitly known. Additionally, using natural food chains allowed us to directly contrast observations of controlled lab raised food chains with our analogous wild collected food chain in order to interpret and ensure the relevance of our laboratory findings for drawing conclusions and lessons that could inform future studies using wild systems.

For laboratory food chains, grain aphids (Sitobion avenae) were raised on two independent food plants; One based on a C3 photosynthetic pathway (wheat Triticum aestivum) and the other based on a C4 pathway (maize Zea mays), enabling separation of the plants on a δ13C axis and thus broadening the generality of any observed patterns. Plants were raised on a common source of homogenised compost and distilled water, and introduced to aphids at 20 days (wheat) or 30 days (maize). 1st generation larvae of wild-caught hoverfly (Syrphus vitripennis) were fed either wheat or maize raised aphids or an approximate 50∶50 ratio of both. Within each treatment, 24–48 hours after hatching, a random subset of hoverfly larvae were exposed to wild caught adult female parasitoids (Diplazon laetorius) to allow parasitic oviposition. All plants and insects were raised under a 16∶8 light∶ dark cycle at 70% humidity. Plant leaves and aphids of all ages were collected at random and frozen (−20°C) prior to tissue preparation. Hoverfly larvae entered pupation 8–10 days after hatching and after 72 hours pupation were frozen (−20°C) for later dissection. Prior experimentation identified 72 hour pupation as suitable to provide soft pupa tissue comparable to that likely consumed by parasitoid larvae (i.e. after exuviae had formed). Parasitised hoverfly larvae were allowed to complete pupae development (19–21 days) and newly eclosed adult parasitoids were frozen (−20°C) within 12 hours without having fed.

For the wild food chain, nettle (Urtica dioica) leaves and nettle aphids (Microlophium carnosum) were collected independently and frozen (−20°C) for later preparation. Hoverfly larvae (Syrphus vitripennis) were collected when judged at >50% grown and laboratory raised on daily-collected wild nettle aphids until pupation, under a 16∶8 light∶dark cycle at 70% humidity. Pupation proceeded until either adult hoverflies or adult parasitoids (Diplazon laetorius) eclosed after 10–11 or 16–20 days respectively, and were frozen (−20°C) within 12 hours without having fed. Our four replicate food chains are hereafter termed after their plants as wheat, maize, wheat + maize (w+m) and nettle. Parasitoid hymenoptera are referred to as wasp hereafter.

Tissue Preparation & Lipid Extraction

We used tissues for sources to best represent assimilation in consumers, as shown to be important [31]. While aphids fed on plant sap, we used whole leaf tissue given difficulties of extracting sap and because whole leaf tissue δ15N has been shown not to differ from sap [32]. For each food chain, following dissection, soft internal tissues of 60–80 aphids were pooled to produce a single sample. Hoverfly larvae soft tissue was obtained from pupae casing. Wasps represented end consumers and we used whole tissues. For wild hoverflies and wasps we used adult whole tissues. Individual hoverflies and wasps each provided single replicates. Sample sizes for plants, aphids and hoverflies were n = 15 (except nettle hoverflies, n = 7) while wasps were more difficult to obtain: wheat (n = 10), w+m (n = 7), nettle (n = 6). For maize, no wasps were obtained due to high larval mortality. All samples were dried at 45°C for >48hrs and homogenised. Subsequently, insect samples were immersed in 2∶1 chloroform: methanol solution for 50 minutes to remove free lipid, and then left to air dry. Prior experimentation showed lipid extraction to have no significant effects on δ15N of these insects.

Stable Isotope Analysis

For all samples, 0.5 mg±0.05 (insect) or 3 mg±0.1 (plant) dried material was enclosed in tin capsules. Stable isotope analysis (SIA) was conducted at the Food and Environment Research Agency, York, UK. Samples were analysed for δ15N and δ13C in a Fisons EA1108 elemental analyser (Carlo Erba Instruments, Milan, Italy), coupled with an Isoprime isotope ratio mass spectrometer (GV Instruments, Manchester, UK). Stable isotope ratios are reported in delta (δ) notation where δ15N and δ13C  =  [(Rsample/Rstandard) −1] ×1000, where R is 15N/14N or 13C/12C. Isotope ratios are expressed in per mil (‰) relative to the ratio of international reference standards (Rstandard) which are Atmospheric Nitrogen and Vienna PeeDee Belemnite (VPDB) for nitrogen and carbon respectively. Measures of standards placed throughout samples exhibited acceptable instrument reproducibility of <0.09‰ (SD) for δ15N and <0.18‰ (SD) for δ13C using collagen standard, insect whole tissue standard (cockroach; Nauphoeta cinerea), and sucrose C4 plant standard.

Data Analysis

Initial analyses used two-way analysis of variance (ANOVA) to test effects of the two explanatory variables trophic level (levels  =  plant, aphid, hoverfly or wasp) and food chain type (levels  =  wheat, maize, w+m or nettle) on δ15N and then δ13C. Model simplification used backwards stepwise regression from a maximal model and ANOVA model comparisons to identify non-significant terms for elimination. Homogeneity of variances and normality of model residuals were checked in all instances. To determine where significant differences lay between levels within treatments, subsequent one-way ANOVA for each food chain type were tested with Tukey post hoc tests, for both δ15N and δ13C.

Within each food chain type, we then calculated mean (± SD) Δδ15N and Δδ13C as the difference in δ15N or δ13C between each source and its consumer by randomly pairing replicates (n = 6 to 15). To establish underlying sources of variation in Δδ15N, two-way ANOVA tested effects of explanatory variables trophic link (levels  =  plant-aphid, aphid-hoverfly, or hoverfly-wasp) and food chain type (levels  =  wheat and nettle). Only nettle and wheat food chain data were used as these contained all three trophic links, allowing for a balanced analysis. For explanatory variables, variation in Δδ15N was quantified using sums of squares in model outputs and was expressed as a proportion of the null model variance. Model simplification was used as described above. All analyses were conducted in R version 2.14.1 [33].

Nitrogen range (NR) and carbon range (CR), calculated as the mean difference between trophic levels of maximum and minimum δ15N or δ13C respectively, are quantifications of trophic structure with NR representing food chain length and CR the breadth of energy sources. For all species combinations of each food chain length within each food chain type separately, NR and CR were independently calculated using resampling of uncertainty around sample mean estimates to provide probabilistic distributions representing 50%, 75% and 95% credible intervals of mean estimates for NR and CR, using the SIBER computational code [20] in the R package SIAR [34], [35]. Additionally for NR, for each food chain length, resampled mean estimates across food chain types and species combinations were pooled to produce overall 95% credible intervals.

As established by preliminary experiments, we applied correction factors of –0.7‰ for δ15N and +0.4‰ for δ13C to hoverflies on the wild nettle food chain (which used adult tissues) to make them directly comparable to larval hoverflies on laboratory food chains.

Data Availability

All raw data used in this study are provided as supplementary material (Dataset S1).

Results

How consistent are Δδ15N and Δδ13C between trophic levels and among food chains?

δ15N.

Though given the same nitrogen source, δ15N of wheat plants were depleted relative to maize plants by ≈2‰, while wild nettle plants (of independent nitrogen source) were slightly enriched (0.3‰) relative to maize. Such differences between food chain types were largely propagated to higher trophic levels (Fig. 1a) and found to be significant; two-way ANOVA showed a significant interaction between trophic level and food chain type indicating effects of these variables were interdependent (F(6, 152)  = 11.57, P<0.001). To determine if large variances in δ15N of maize and wheat plants (Fig. 1a) disproportionately affected these results, we repeated this analysis excluding these plants, but found no difference in outcome.

thumbnail
Download:
Figure 1. δ15N (a) and δ13C (b) across four trophic levels of four replicate terrestrial food chains.

Mean ± SD (‰) are shown. n = 6 to 15. For plant δ15N, mean ± SD are offset on x-axis for clarity. Dotted lines are trophic links between two aphid prey sources and their hoverfly predator.

https://doi.org/10.1371/journal.pone.0093281.g001

To ascertain patterns of δ15N discrimination, Tukey post hoc tests performed on one–way ANOVA for each food chain type established where δ15N differed between trophic levels. Overall, δ15N discrimination factors showed significant enrichment from source to consumer (range of 1.4‰ to 3.3‰), in all but three instances (Table 1). Exceptional to this trend, wheat feeding aphids showed significant average depletion in δ15N (−2.4‰) relative to hosts.

thumbnail
Download:
Table 1. Mean ± SD (‰) Δδ15N and Δδ13C across four trophic levels of four terrestrial food chains.

https://doi.org/10.1371/journal.pone.0093281.t001

Given δ15N enrichment was broadly consistent and larger than that observed in δ13C (Table 1), for δ15N we also used two-way ANOVA to determine sources of variation in Δδ15N (Table 2). After disproportionate plant variation was excluded, we observed a significant interaction between trophic link and food chain type accounting for 32% of variation in Δδ15N. This significant interaction infers variation in Δδ15N as caused by these variables was species and system specific.

thumbnail
Download:
Table 2. Sources of variance in Δδ15N (‰).

https://doi.org/10.1371/journal.pone.0093281.t002

δ13C.

δ13C values were ≈17‰ different between wheat and maize food chains, with hoverfly and wasps on the w+m food chain approximately half way between the two having integrated aphid sources from both (Fig. 1b). Two-way ANOVA showed δ13C was significantly affected by an interaction between trophic level and food chain type (F(6, 152)  = 19.22, P<0.001), indicating food chain effects on δ13C were affected inconsistently by trophic level (Fig. 1b).

Differences in δ13C discrimination across trophic links and food chain types were of variable direction and magnitude (−0.7‰ to 1.9‰, excluding w+m aphid-hoverfly; Table 1). Across trophic links and food chain types, Tukey post hoc analyses showed either significant enrichment (0.6‰ to 1.9‰) or no change in δ13C between trophic levels (Table 1).

Does nitrogen range accurately determine food chain length?

In most instances, estimates of nitrogen range (NR) were observed to increase with greater food chain length within each food chain (Fig. 2). Exceptional to this were low NR estimates on the wheat and w+m food chains for combinations including wheat plants.

thumbnail
Download:
Figure 2. Probability distributions of mean nitrogen range (NR) for different food chain lengths for each food chain.

NR is difference between mean δ15N (‰) of community end members, based on resampling (n = 10,000) of uncertainty in sample mean estimates. Black dots represent mode (of means), while shaded boxes (dark to light) show 50%, 75% and 95% credible intervals for mean estimates. No wasps were obtained for the maize food chain. x-axis labels are species identity: p = plant, a = aphid, h = hoverfly, w = wasp. Parenthesis number shows food chain length.

https://doi.org/10.1371/journal.pone.0093281.g002

Within a food chain, in some instances modal or credible interval values of NR varied distinctly between different species combinations of the same food chain length, but such differences were not constant across food chains suggesting a species and system specific nature to such variation. Across food chains, 95% credible interval estimates of mean NR were not generally larger than 1‰ δ15N either side of the modal value (Fig. 2). The exception was larger estimates on the wheat food chain when wheat plants were included as a consequence of uncertainty in mean estimates (Fig. 2), propagated from large sample variation in wheat plants (Fig. 1).

Overall estimates of NR based on combining resampled mean estimates from all four replicate food chains and all species combinations (for each level of food chain length), showed modal NR values to increase by between 1.2‰ and 2.7‰ with each additional trophic level (Fig. 3). Overlap in 95% credible intervals between different food chain lengths (Fig. 3a) was reduced when wheat plant combinations were excluded (Fig. 3b), and then further reduced when maize plant combinations were also excluded (Fig. 3c). Notably, there was no subsequent change in 95% credible intervals when nettle plants were additionally removed (Fig. 3d), suggesting that these wild plants did not contribute noticeably to variation in estimates of NR. Following removal of wheat and maize plant combinations, increase in modal NR with trophic level was more closely matched at 1.7‰ and 2.4‰. Removal of nettle plant combinations increased modal values, most likely due to nettle-aphid discrimination factors being lower than some other trophic links (Table 1).

thumbnail
Download:
Figure 3. Overall nitrogen range (NR) for food chains containing two, three or four trophic levels.

Values derived from combining resampled mean values from all four replicate food chains. TL  =  trophic levels. Circles represent mode (of means) and bars 95% credible intervals for mean. a) all combinations: 2 TL n = 10, 3 TL n = 7, 4 TL n = 3; b) excludes combinations that include wheat plants: 2 TL n = 9, 3 TL n = 5, 4 TL n = 1; c) excludes all combinations that include wheat or maize plants: 2 TL n = 8, 3 TL n = 4, 4 TL n = 1; d) excludes all combinations containing primary producers: 2 TL n = 7, 3 TL = 3.

https://doi.org/10.1371/journal.pone.0093281.g003

How does carbon range change with food chain length?

For estimates of carbon range (CR), we observed few consistent patterns in CR across differing food chain lengths within or among food chains (Fig. 4). Wheat and maize food chains both showed no pattern of change in CR with changing food chain length; wheat CR was <1‰ across two, three and four trophic levels, while maize CR was 2‰ for combinations of both two and three trophic levels. For the w+m food chain there was marginal increase in modal CR with food chain length (two to four trophic levels ≈0.3‰ to 2.3‰) but this was considerably less than variation in mean estimates as shown by large 95% credible intervals (>3‰), on account of two isotopically disparate δ13C plant sources. Conversely, the wild nettle food chain trended towards greater modal CR with food chain length (two to four trophic levels ≈1‰ to 3.5‰); however, overlap in 95% credible intervals between different food chain lengths was observed, while modal values of different species combinations of the same food chain length also differed by ≈1‰ to 1.5‰.

thumbnail
Download:
Figure 4. Probability distributions of mean carbon range (CR) for different food chain lengths within each food chain.

Distributions based on resampling (n = 10,000) of uncertainty in sample mean estimates. 50%, 75% and 95% credible intervals as in Fig. 2. x-axis: p = plant, a = aphid, h = hoverfly, w = wasp. Numbers in parenthesis show food chain length.

https://doi.org/10.1371/journal.pone.0093281.g004

Discussion

We tested the accuracy of isotopic measures of trophic structure against known trophic positions using well replicated terrestrial plant and invertebrate food chains spanning four trophic levels. We found that despite some overlap in nitrogen range (NR) between longer food chains, across a range of different species combinations and food chains NR generally quantified food chain length well, suggesting robustness to observed variation in discrimination. Additionally, we found few consistent trends in δ13C discrimination with typically small (0.6‰ to 1.9‰) or no enrichment, and concurrently little and inconsistent change in CR with food chain length, emphasising the utility of δ13C to trace diet–consumer pathways. We suggest our estimates of food chain trophic structure are particularly robust because they were calculated using resampling procedures, allowing for propagation of variance in Δδ15N and Δδ13C into quantified uncertainty in final structural estimates.

How consistent are Δδ15N and Δδ13C between trophic levels and among food chains?

δ15N.

Within species variation in δ15N of 0.3‰ to 0.9‰ (excluding plants) was observed, and was a likely consequence of differences in Δδ15N between individuals due to compositional differences in food consumed and metabolic differences in assimilation of that food [13], [27], [36] (and references therein). An additional anomalous observation was large δ15N variation in wheat and maize plants that was not observed in wild nettle plants, suggesting that laboratory conditions affected variation. As such variation was not observed for δ13C which is sourced from the atmosphere and showed much smaller comparable variation across wheat, maize and nettle plants, we speculate δ15N variation was explained by micronutrient pockets in laboratory soil medium through incomplete homogenisation. Alternately, stress, as caused by unnaturally high and variable laboratory aphid density, may have affected plant metabolism and hence nitrogen balance, as shown for other taxa [37], [38]. As we pooled aphids to produce each aphid sample, plant δ15N variation likely averaged across these multiple aphids, explaining why large δ15N variation was not subsequently seen in aphids.

We found significant δ15N enrichment from source to consumer in a majority of instances, varying between 1.4‰ and 3.3‰, a range concurrent with literature estimates [11]–[13]. Averaged across food chains, such enrichment was 2.2‰ (excluding discrimination from wheat and maize plants to their aphids). This value is marginally lower than literature average estimates for invertebrates of 2.5‰ [12], [13], and lower than the overall literature estimate of 3.4‰ which is commonly employed to calculate trophic levels and food chains (e.g. [11], [23], [24]). This higher overall literature value may be reflective of trophic discrimination factors associated with vertebrates [13] and thus the lower enrichments noted in our insects re-emphasises the need to use taxa-specific discrimination factors when using isotopic data to calculate ecological parameters [13]. Additionally, broad enrichment range of 1.9‰ (1.4‰ to 3.3‰) in our results, as concurrent with previous studies [11]–[13], demonstrates variation in trophic discrimination factors is an important artefact in isotopic data that should be accounted for. Thus the use of averaging to produce commonly used trophic discrimination factors excludes variation in discrimination from final ecological estimates. Given such variation was present in our system, our subsequent use of resampling procedures allowed us to propagate discrimination uncertainty into our estimates of food chain structure, producing estimates that we contend are more accurate and hence ecologically robust [20].

In this study, as a single exception in source to consumer enrichment, wheat feeding aphids were on average depleted in δ15N relative to hosts, concurring with other studies [32], [39]. A previously observed negative relationship between aphid δ15N discrimination and host plant total nitrogen content [32] is consistent with our results; aphid enrichment was observed on maize and nettle plants that had low total nitrogen contents of 1.7% and 2.7% respectively, while wheat plants had higher nitrogen of 5.4%.

While examining sources of variation in Δδ15N, we found a significant interaction between trophic link and food chain type, suggesting effects of these variables were inconsistent and thus could not be generalised. This result is in close agreement with recent observations of discrimination variation at broad spatial scales in wild systems whereby variation was dissimilar across taxa within and between trophic groupings [28]. Mechanistic underpinnings explaining variation in Δδ15N are beyond the scope of this study but briefly, for different trophic links and for the same trophic links on different food chains, variation in Δδ15N was likely due to differences in nitrogen [32] and protein [40] quality of food, and/or consumer's differing metabolic assimilation of that food [27] (and references therein).

δ13C.

Trophic discrimination of δ13C from source to consumer was inconsistent between trophic links and food chain type, ranging from −0.7‰ to 1.9‰, and more generally observed as either showing no change or significant enrichment of between 0.6‰ to 1.9‰, concurring with discrimination reviews [11], [13]. Given that in our system the carbon axis was broad, such inconsistent and generally small trophic discriminations meant δ13C was diagnostic of food chain type. Similarly, δ13C signatures of hoverfly predators on the w+m food chain were intermediate of their two disparate aphid sources, illustrating well the usefulness of δ13C data to integrate and reflect dietary sources.

Does nitrogen range accurately determine food chain length?

Using resampling procedures to calculate nitrogen range (NR), we tested how NR changed with known food chain length within food chains, and then across food chains types. Excluding wheat plant–aphid combinations, we found that NR accurately determined food chain length within all food chains. This finding is concurrent with a previous validation showing positive correlations between trophic positions of freshwater fish estimated by both δ15N and traditional gut content analysis [14].

In our study, the inclusion of wheat plant–aphid combinations depressed NR measures on the wheat and w+m food chain types. This was because aphids were depleted relative to wheat plants, such that aphids were effectively base of the δ15N food chain while wheat constituted an additional trophic level that did not act to extend NR. This importantly shows exceptional species-specific effects may adversely affect the accuracy of isotopic measures of food chain length. As previously we had identified the wheat plant–aphid relationship as an exception to the generic enrichment in δ15N from source to consumer, we feel justified in concluding that, more generally, NR predicted food chain length.

Overall, when all trophic combinations for all food chain types were combined, modal NR values increased by between 1.2‰ and 2.7‰ with each trophic level (or 1.7‰ and 2.4‰ excluding wheat and maize plant combinations), suggesting a robustness of this technique for calculating food chain length. Excluding combinations including wheat and maize plants (on account of large variation in their δ15N signatures), 95% credible intervals of NR estimates showed some overlap between food chain lengths of three and four trophic levels, inhibiting accurate estimation of food chain length at these points of overlap. Similarly, for wild-sampled freshwater algae, invertebrate and fish food chains it has been shown that differences in δ15N between trophic levels was largely reflective of food chain length, though with variation in Δδ15N causing some uncertainty/error in estimating trophic level [28]. Current practice uses NR to estimate food chain length and determine subsequent conclusions, but rarely has NR's use been tested, with terrestrial systems particularly understudied [27]. Based on our results, we urge caution interpreting food chain length when NR values fall in known overlap boundaries; i.e. in our study NR values of 5‰ or 6‰ could be either three or four trophic levels. Such overlap was a consequence of variation in NR values within each trophic link and between trophic links of the same food chain length, both within and between food chain types, as ultimately caused by variation in Δδ15N within and between source-consumer pairings (Table 1). Of the few studies that have examined effects of variation in baseline δ15N [17] and Δδ15N [11], [25], [28] on error of trophic level or food chain length estimates, all have found error concurrent to the variation we show in estimates of NR. Importantly, our research diversifies these studies by testing empirical rather than mainly theoretical measures of food chains, in a controlled laboratory context, using understudied terrestrial and invertebrate systems, in addition to providing robust estimates of NR based on resampling of variation.

Additionally, as different food chain types in our experiment differed in their NR values for given food chain lengths as a consequence of variation in Δδ15N, so it is worth urging some caution when using NR for direct comparisons between systems of simple single-pathway food chains. It has been reasonably suggested that such variation may average out over multiple trophic levels or larger sample sizes [27], though based on our results we would call for research to further test the importance of such variation. Pragmatically, as in our study the effect of food chain type on Δδ15N was variable, then studies sampling across multiple food chains will absorb such variation as ‘noise’ that may be acceptable or specifically filtered out if food chain identity was categorised in analyses as a random effect. We speculate that such variation may also be averaged out when considering overall structure of larger food webs containing multiple food chain pathways.

Our overall estimates of NR were based on exact trophic positions in well replicated food chains and used resampling procedures that propagated uncertainty in discrimination factors to produce estimates that represented a full range of possible NR values. Given such an approach is likely to produce robust estimates of NR, and that more generally these estimates suggested that NR can accurately discriminate different food chain lengths, we conclude NR is a useful isotopic metric for quantifying food chain length.

How does carbon range change with food chain length?

Using resampling procedures we calculated carbon range (CR) values for all food chain length combinations on all four food chain types. Overall, our results show that across trophic levels CR changed little and inconsistently both within and between food chain types (Fig. 4). Little change in CR over four trophic levels (modal range across food chain types <1‰ to 3.5‰; Fig.4) suggests fidelity of δ13C values between primary producers and top predators, as concurrent with our earlier findings of small and inconsistent δ13C discrimination (Table 1). Given global estimates of aquatic food chains approximate 3.5 to 4.0 trophic levels [17] little change in CR over four trophic levels in our study demonstrates more broadly the utility of δ13C to trace energy pathways, be they either simple singular chains or potentially when embedded within larger trophic structures. However, a potential caveat of using δ13C is that source signatures may need to be distinct in order to determine different energy pathways; we acknowledge that our δ13C axis was broad (≈17‰) with different plant species showing separation in δ13C values. In systems where multiple consumers feed on a single basal source, or basal sources have closely aligned δ13C, identification of different energy pathways may not be possible.

Stable isotope approaches to quantifying food web structure continue to proliferate in ecological studies and so improving our understanding of the accuracy and limitations of these techniques is of importance. In this study, source-consumer δ15N discrimination generally showed enrichment, with broad variation (range 1.9‰) between trophic levels and among food chains which propagated variation to subsequent estimates of NR. However, across a range of species combinations and food chains we show NR proved robust to such variation, distinguishing food chain length well, though some overlap between longer food chains importantly establishes limits in NR's precision. CR changed little with food chain length and hence δ13C is potentially a useful tracer of source-consumer interactions. Having established that variation in discrimination affected estimates of trophic structure, we recommend the use of resampling procedures to propagate variation as quantified uncertainty in final estimates of structure. Such procedures are necessary to improve accuracy and robustness of ecological conclusions in future isotopic studies. Given global estimates of aquatic food chains approximate four trophic levels, and that most food chains include invertebrates, our use of four trophic level plant-invertebrate food chains makes our findings relevant to a majority of ecological systems and contexts.

Supporting Information

Dataset S1.

δ15N and δ13C values for all plant and invertebrates samples for all four food chains used in this research. All invertebrate samples underwent lipid extraction prior to stable isotope analysis and thus δ15N and δ13C values given are post lipid extraction.

https://doi.org/10.1371/journal.pone.0093281.s001

(XLSX)

Acknowledgments

We thank Anna Cucknell for laboratory expertise, Francis Gilbert and the hoverfly helplist (syrphidae@lists.nottingham.ac.uk) for advice, and two anonymous reviewers for comments.

Author Contributions

Conceived and designed the experiments: SB RM FVV MP. Performed the experiments: MP SK GR. Analyzed the data: MP SB. Contributed reagents/materials/analysis tools: SK GR FVV. Wrote the paper: MP SB.

References

  1. 1. Doi H, Vander Zanden MJ, Hillebrand H (2012) Shorter food chain length in ancient lakes: evidence from a global synthesis. PLoS One 7: e37856.
  2. 2. Price SA, Hopkins SSB, Smith KK, Roth VL (2012) Tempo of trophic evolution and its impact on mammalian diversification. Proc Natl Acad Sci USA 109: 7008–7012.
  3. 3. Thompson RM, Townsend CR (2005) Energy availability, spatial heterogeneity and ecosystem size predict food-web structure in streams. Oikos 108: 137–148.
  4. 4. Srivastava DS, Bell T (2009) Reducing horizontal and vertical diversity in a foodweb triggers extinctions and impacts functions. Ecol Lett 12: 1016–1028.
  5. 5. Rooney N, McCann KS (2012) Integrating food web diversity, structure and stability. Trends Ecol Evol 27: 40–46.
  6. 6. Cardinale BJ (2011) Biodiversity improves water quality through niche partitioning. Nature 472: 86–91.
  7. 7. Thompson RM, Brose U, Dunne JA, Hall RO Jr, Hladyz S, et al. (2012) Food webs: reconciling the structure and function of biodiversity. Trends Ecol Evol 27: 689–697.
  8. 8. Bearhop S, Adams CE, Waldron S, Fuller RA, Macleod H (2004) Determining trophic niche width: a novel approach using stable isotope analysis. J Anim Ecol 73: 1007–1012.
  9. 9. DeNiro MJ, Epstein S (1978) Influence of diet on the distribution of carbon isotopes in animals. Geochim Cosmochim Acta 42: 495–506.
  10. 10. DeNiro MJ, Epstein S (1981) Influence of diet on the distribution of nitrogen isotopes in animals. Geochim Cosmochim Acta 45: 341–351.
  11. 11. Post DM (2002) Using stable isotopes to estimate trophic position: models, methods, and assumptions. Ecology 83: 703–718.
  12. 12. Vanderklift MA, Ponsard S (2003) Sources of variation in consumer - diet δ15N enrichment: a meta – analysis. Oecologia 136: 169–182.
  13. 13. Caut S, Angulo E, Courchamp F (2009) Variation in discrimination factors (δ15N and δ13C): the effect of diet isotopic values and applications for diet reconstruction. J Appl Ecol 46: 443–453.
  14. 14. Vander Zanden MJ, Cabana G, Rasmussen JB (1997) Comparing trophic position of freshwater fish calculated using stable nitrogen isotope ratios (δ15N) and literature dietary data. Can J Fish Aquat Sci 54: 1142–1158.
  15. 15. Vander Zanden MJ, Rasmussen JB (1999) Primary consumer δ13C and δ15N and the trophic position of aquatic consumers. Ecology 80: 1395–1404.
  16. 16. Cabana G, Rasmussen JB (1996) Comparison of aquatic food chains using nitrogen isotopes. Proc Natl Acad Sci USA 93: 10844–10847.
  17. 17. Vander Zanden MJ, Fetzer WW (2007) Global patterns of aquatic food chain length. Oikos 116: 1378–1388.
  18. 18. Layman CA, Arrington DA, Montana CG, Post DM (2007) Can stable isotope ratios provide for community - wide measures of trophic structure? Ecology 88: 42–48.
  19. 19. Schmidt SN, Olden JD, Solomon CT, Vander Zanden MJ (2007) Quantitative approaches to the analysis of stable isotope food web data. Ecology 88: 2793–2802.
  20. 20. Jackson AL, Inger R, Parnell AC, Bearhop S (2011) Comparing isotopic niche widths among and within communities: SIBER – Stable Isotope Bayesian Ellipses in R. J Anim Ecol 80: 595–602.
  21. 21. Layman CA, Araujo MS, Boucek R, Hammerschlag-Peyer CM, Harrison E, et al. (2011) Applying stable isotopes to examine food – web structure: an overview of analytical tools. Biol Rev Camb Philos Soc 87: 545–562.
  22. 22. Post DM, Pace ML, Hairston NG (2000) Ecosystem size determines food-chain length in lakes. Nature 405: 1047–1049.
  23. 23. Takimoto G, Spiller DA, Post DM (2008) Ecosystem size, but not disturbance, determines food-chain length on islands of the Bahamas. Ecology 89: 3001–3007.
  24. 24. McHugh PA, McIntosh AR, Jellyman PG (2010) Dual influences of ecosystem size and disturbance on food chain length in streams. Ecol Lett 13: 881–890.
  25. 25. Vander Zanden MJ, Rasmussen JB (2001) Variation in δ15N and δ13C trophic fractionation: Implications for aquatic food web studies. Limnol Oceanogr 46: 2061–2066.
  26. 26. Matthews B, Mazumder A (2004) A critical evaluation of intrapopulation variation of δ13C and isotopic evidence of individual specialization. Oecologia 140: 361–371.
  27. 27. Martinez del Rio C, Wolf N, Carleton SA, Gannes LZ (2009) Isotopic ecology ten years after a call for more laboratory experiments. Biol Rev Camb Philos Soc 84: 91–111.
  28. 28. Bunn SE, Leigh C, Jardine TD (2013) Diet-tissue fractionation of δ15N by consumers from streams and rivers. Limnol Oceanogr 58: 765–773.
  29. 29. McCutchan JH, Lewis WM, Kendall C, McGrath CC (2003) Variation in trophic shift for stable isotope ratios of carbon, nitrogen, and sulfur. Oikos 102: 378–390.
  30. 30. Jardine TD, Kidd KA, Fisk AT (2006) Applications, considerations, and sources of uncertainty when using stable isotope analysis in ecotoxicology. Environ Sci Technol 40: 7501–7508.
  31. 31. Perkins MJ, McDonald RA, van Veen FJF, Kelly SD, Rees G, et al. (2013) Important impacts of tissue selection and lipid extraction on ecological parameters derived from stable isotope ratios. Methods Ecol Evol 4: 944–953.
  32. 32. Wilson ACC, Sternberg LSL, Hurley KB (2011) Aphids alter host-plant nitrogen isotope fractionation. Proc Natl Acad Sci USA 108: 10220–10224.
  33. 33. R Development Core Team (2011) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. Available: http://www.R-project.org/.
    • 34. Parnell AC, Inger R, Bearhop S, Jackson AL (2008) SIAR: Stable isotope analysis in R.
      • 35. Parnell AC, Inger R, Bearhop S, Jackson AL (2010) Source partitioning using stable isotopes: coping with too much variation. PLoS One 5: e9672.
      • 36. Boecklen WJ, Yarnes CT, Cook BA, James AC (2011) On the use of stable isotopes in trophic ecology. Annu Rev Ecol Evol Syst 42: 411–440.
      • 37. Hobson KA, Alisauskas RT, Clark RG (1993) Stable – nitrogen isotope enrichment in avian tissues due to fasting and nutritional stress: implications for isotopic analyses of diet. Condor 95: 388–394.
      • 38. Voigt CC, Matt F (2004) Nitrogen stress causes unpredictable enrichments of 15N in two nectar-feeding bat species. Journal Exp Biol 207: 1741–1748.
      • 39. Oelbermann K, Scheu S (2002) Stable isotope enrichment (δ15N and δ13C) in a generalist predator (Pardosa lugubris, Araneae: Lycosidae): effects of prey quality. Oecologia 130: 337–344.
      • 40. Robbins CT, Felicetti LA, Sponheimer M (2005) The effect of dietary protein quality on nitrogen isotope discrimination in mammals and birds. Oecologia 144: 534–540.