Clarifying a trophic black box: stable isotope analysis reveals unexpected dietary variation in the Peruvian anchovy Engraulis ringens

Introduction

Boundary current ecosystems such as the Humboldt Current System (HCS) are characterised by high biological productivity driven by coastal upwelling of cold sub-surface nutrient-rich waters (Chavez & Messié, 2009). Food chains in such ecosystems have been typically considered as simple and short (Ryther, 1969), with high efficiency of trophic transfer between trophic levels. Here, phytoplankton form the base of the food web, and are consumed by zooplankton and small-bodied pelagic fishes such as the Peruvian anchovy Engraulis ringens Jenyns, 1842 (from hereon in, anchovy). The small pelagic fish assemblage is typically dominated by one or a few species (Bakun, 1996), which due to their sheer abundance and biomass can exercise control over the trophic dynamics of the whole ecosystem (Cury et al., 2000).

Biological production in the HCS (Herrera & Escribano, 2006) is such that it supports the capture of more fish per unit area than any other environment in the world (Chavez et al., 2008). Indeed, production in the HCS is considered anomalous, even among eastern current systems (Bakun & Weeks, 2008; Chavez & Messié, 2009). Industrial fisheries in Peru and Chile target anchovies and other pelagic fishes such as jack mackerel Trachurus murphyi and sardines (Sardinops sagax). The anchovy stock in the region is the most heavily exploited fish worldwide: although annual catches vary considerably reflecting the marked effects of ENSO on stock size, between 1990 and 2015, on average (median ± IQR) 7,419,295 ± 3,329,637 tonnes of anchovies were removed per year (FAO, 2018).

Although anchovy (and other small pelagics) support globally important fisheries (Chavez & Messié, 2009; Kämpf & Chapman, 2016), and play integral roles in structuring upwelling ecosystems (Cury et al., 2000), many basic aspects of their ecology are understudied. For instance, information on their trophic ecology is limited, and likely constrains our understanding of their role in the HCS, and the utility of ecosystem model outputs. Studies of anchovy diet based on stomach content analysis (SCA) of juveniles and adults typically report that stomach contents are dominated (in numerical terms) by phytoplankton (Espinoza & Bertrand, 2008; Medina et al., 2015; Ryther, 1969), and anchovies have been classically considered as phytoplanktivores. This low trophic position has been proposed as an explanation for their great abundance in the HCS, and the ecosystem’s anomalous capacity to support such high production of pelagic fishes.

Most studies of anchovy diet in the literature are based on counts of individual stomach contents, which is known to bias conclusions when prey of markedly different sizes are consumed (Hyslop, 1980). Not surprisingly, microscopic food items such as diatoms can be extremely abundant, leading them to dominate relative estimates of prey importance (up to 98%: Whitehead, Nelson & Wongratana, 1998), even though zooplankton are commonly reported from anchovy stomach contents. Given the putative importance of phytoplankton to their diet, many researchers have assumed that anchovy have a trophic position of ca. 2–2.5 (Guénette, Christensen & Pauly, 2008; Pauly et al., 1998). These values are commonly used in ecosystem models such as Ecopath, and anchovies are considered a classic low-trophic level species (Smith et al., 2011). This estimate of anchovy TP has played an important role in one of the largest controversies in modern fisheries science—the issue of global fishing down the food web (Pauly et al., 1998; Pauly, Froese & Palomares, 2000). The global importance of the HCS anchovy fishery is such that if its statistics are included, it skews global estimates of the mean trophic level of catch downward.

More recently however, there has been an important reassessment of anchovy diet. Espinoza & Bertrand (2008), rather than counting stomach contents, focused on the energetic content of prey, and clearly showed that zooplankton (mainly copepods and euphausiids) were the principle source of energy to anchovies. Beyond highlighting a key issue of using counts of stomach contents to estimate TP, their results had major implications regarding our understanding of how this globally important marine ecosystem functions (e.g.,  Ballón et al., 2011). Given the biomass of anchovies in the HCS, and their dominant role in the food web, changes in our understanding of their putative diet has subsequent impacts on how we interpret the movement of energy and nutrients through the food web, with consequences for qualitative and quantitative models of HCS function and resource management.

SCA has long been the gold standard for assessing what fish consume (Hynes, 1950; Hyslop, 1980), but clearly can bias our understanding of how fish direct the flow of materials through a food web (Espinoza & Bertrand, 2008; Hyslop, 1980). An alternative to the snapshot of recently ingested prey provided by SCA, is to take a biochemical approach to assessing diet such as stable isotope analysis (SIA) or fatty acid analysis (Nielsen et al., 2018). The advantages of SIA in particular, are that the technique provides information on prey assimilation over longer temporal scales than SCA (weeks–months), with the period depending on the tissue sampled (Thomas & Crowther, 2015). By combining analysis of carbon and nitrogen stable isotope ratios of consumers and their putative prey, it is possible to characterise the source of energy and nutrients assimilated by a consumer (Parnell et al., 2013). Furthermore, if isotope values are available for both the consumer and the base of the food web (Vander Zanden & Rasmussen, 1996), it is possible to estimate the long term trophic position at which a consumer feeds (Quezada-Romegialli et al., 2018).

Recent work in central Chile and Peru using SIA has provided further evidence that previous assumptions regarding anchovy trophic position were wrong. Hückstadt, Rojas & Antezana (2007) compared POM and anchovy δ¹⁵N values from central Chile and estimated anchovy TP as 3.6. In a wide-ranging recent study from the northern HCS, Espinoza et al. (2017) compared anchovy (and other taxa) δ¹⁵N values relative to copepod δ¹⁵N to estimate TP. They estimated that anchovy TP ranged between 3.4 and 3.7. Both of these studies provide evidence that anchovy TP is at least a trophic level above values typically used for the species in trophic models (Guénette, Christensen & Pauly, 2008; Pauly et al., 1998), and challenge the concept of anchovies as a low trophic level fish (Smith et al., 2011). Espinoza et al.’s (2017) estimates for anchovy TP are important as they support the authors’ previous findings from SCA indicating that phytoplankton, although abundant numerically in stomachs, did not make a major contribution to assimilated diet. However, their and Hückstadt, Rojas & Antezana’s (2007) estimates were based on a simple model which does not include isotopic variation in trophic discrimination or in the baseline itself (see Quezada-Romegialli et al., 2018 for a description of the issue). The latter point is particularly important in the Espinoza et al. (2017) case, as the use of copepods as a baseline likely introduces considerable error compared to e.g., the use of a primary producer, given the range of trophic strategies displayed by pelagic marine copepods (Giesecke & González, 2004) and the uncertainties in allocating a TP for mixed samples of copepods.

It is becoming increasingly apparent that food webs associated with upwelling ecosystems can be more complex and dynamic than previously thought (Docmac et al., 2017; Espinoza & Bertrand, 2008). As such, there is a need for improved understanding of how these systems function, in order to inform and update existing trophic models used to explain the flow of energy and nutrients, as well as to allow an informed management of a globally important fishery. Here, we examine the trophic ecology of adult anchovies from northern Chile using stable isotope ratios of anchovies and their putative prey to estimate the relative role of phytoplankton and other prey, and to provide robust estimates of anchovy trophic position. Given the debate over the trophic ecology of anchovy, we compared their stable isotope values with that of juvenile jack mackerel, a known pelagic carnivore (Alegre et al., 2015; Orrego & Mendo, 2015).

Materials & Methods

Study area

Samples were collected between 20°30′S and 21°30′S off the coast of northern Chile (Fig. 1) during the Austral winter of 2008. This area is characterized for having persistent winds that permit year-round upwelling (although upwelling often strengthens during Spring–Summer), generating the intrusion of cold nutrient rich sub-surface waters along the shore (Thiel et al., 2007).

During the study period, neutral environmental conditions were present (i.e., non-ENSO). Sea surface temperatures ranged between 15.2 and 17.5 °C, salinities varied between 34.6 and 34.9 and surface dissolved oxygen concentrations between 4.2 and 6.7 ml l⁻¹. In general terms, the study area was characterized by a low-magnitude permanent upwelling, more marked around a latitude of 21°10′S (Fuenzalida et al., 2009).

Results

Isotopic values for the different taxa analysed are shown in Fig. 2. Phytoplankton had a mean δ¹³C value of −18.6‰ but showed considerable variation between sampling locations (SD = 1.6‰). Phytoplankton mean δ¹⁵N was 12.4 (± 1.3‰), reflecting the influence of ¹⁵N-enriched upwelling-derived NO₃, and apparently formed the basal resource for consumers from overlying trophic levels. Carbon stable isotope values for the different putative zooplankton prey classes (anchovy eggs and larvae, unidentified fish eggs, copepods, crustacean larvae, euphausiids) ranged between −20.3 and −17.9‰ for δ¹³C and 15.1 to 19.7‰ for δ¹⁵N.

Carbon and nitrogen stable isotope values were estimated from 30 adult anchovies (28 male, 2 female). Anchovy TL varied between 13.7 and 16.1 cm (mean TL = 14.8 ± 0.6 cm), and blotted wet mass between 13.9 and 25.7 g (19.5 ± 2.9 g). Juvenile jack mackerel (one male, one female, 18 indeterminate) TL varied between 18.5 and 22.2 cm (19.5 ± 1.1 cm), and mass between 43.2 and 79.8 g (53.7 ± 9.8 g).

Anchovy lipid-free δ¹³C values varied between −18.7 and −16.1‰, with a mean of −17.1 ± 0.6‰. Variation in δ¹⁵N values was more marked, with individual anchovies ranging between 13.8 and 22.8‰, and a mean δ¹⁵N of 19.7 (±1.9)‰. Examination of anchovy δ¹⁵N values revealed the presence of two apparent groups differing in their δ¹⁵N values: one group was relatively ¹⁵N enriched (21.6 ± 0.9‰), and the second was made up of ¹⁵N-depleted individuals (18.5 ± 1.3‰).

Jack mackerel lipid-free δ¹³C values ranged between −18.4 and −17.6‰ (mean δ¹³C −18.0 ± 0.2‰), while their δ¹⁵N values ranged between 18.3 and 22.2‰, with a mean δ¹⁵N of 21.6 (± 0.9)‰, apparently similar to that of the high δ¹⁵N group of anchovies.

There was no evidence of correlation between the isotopic composition and mass (Fig. 3) of anchovy (δ¹³C: R_S =  − 0.02, P = 0.94; δ¹⁵N: R_s = 0.11, P = 0.56). Jack mackerel δ¹³C was similar across the size range examined (R_s =  − 0.27, P = 0.27) but showed a positive relationship between mass and δ¹⁵N (δ¹⁵N: R_s = 0.55, P = 0.01). Anchovies were significantly smaller than jack mackerel (Welch’s t test of mass: t_28.1 =  − 15.1, P < 0.001), but there was no evidence for any size difference between the two anchovy groups (t_17.2 = 0.70, P = 0.493).

Mixing models

Mixing models results (Fig. 4, Table 1) indicate that all fish examined were carnivorous, with little evidence for large-scale contributions by phytoplankton. When considered as a single group, anchovies largely assimilated carbon and nitrogen from crustacean larvae (median estimate = 61%) and anchovy larvae (15%), with phytoplankton making a minor contribution (5%). When considered separately, the anchovy group with high δ¹⁵N values had a large contribution of anchovy larvae in their diet (55%), while the low δ¹⁵N anchovy group were estimated to have assimilated the majority of their carbon and nitrogen from crustacean larvae (73%). Mixing model results indicated that anchovy larvae (39%), unidentified fish eggs (19%) and anchovy eggs (15%) all made major contributions to the assimilated diet of jack mackerel. Comparison between the mixing model results for the high δ¹⁵N group of anchovies and jack mackerel indicated that they had broadly similar foraging modes.

Table 1:

Summary statistics (median (credibility limits)) for estimated contribution of different putative prey to the assimilated diet of anchovy (all anchovy combined, low δ¹⁵N and high δ¹⁵N groups) and jack mackerel estimated using SIMMR.

	Estimated proportional contribution to diet
Putative prey	All anchovy	High δ15N anchovy	Low δ15N anchovy	Jack mackerel
Anchovy eggs	0.06 (0.01–0.26)	0.08 (0.01–0.36)	0.04 (0.01–0.15)	0.15 (0.02–0.36)
Anchovy larvae	0.15 (0.02–0.34)	0.55 (0.11–0.74)	0.03 (0.01–0.13)	0.39 (0.18–0.59)
Copepods	0.03 (0.01–0.1)	0.04 (0.01–0.14	0.03 (0.01–0.13)	0.05 (0.01–0.13)
Crustacean larvae	0.61 (0.37–0.77)	0.06 (0.01–0.23)	0.73 (0.42–0.88)	0.09 (0.01–0.23)
Euphausiids	0.02 (0.001–0.08)	0.04 (0.01–0.15)	0.02 (0.004–0.08)	0.05 (0.01–0.14)
Unidentified fish eggs	0.03 (0.001–0.10)	0.09 (0.01–0.44)	0.02 (0.004 –0.07)	0.19 (0.05–0.34)
Phytoplankton	0.05 (0.001–0.22)	0.05 (0.01–0.17)	0.07 (0.01–0.36)	0.06 (0.01–0.18)

DOI: 10.7717/peerj.6968/table-1

Trophic position

Bayesian estimates of TP supported the mixing model results, with credibility limits for all fish including TP 3 (Fig. 5). Modal (95% credibility limits) estimated TP for the pooled anchovy sample was 3.23 (2.93–3.58). The low δ¹⁵N anchovy group had a lower estimated TP (2.91: 2.62–3.23). Estimates of TP for the high δ¹⁵N anchovy group (3.79: 3.48–4.16) and jack mackerel (3.80: 3.51–4.14) were highly similar.

Discussion

Our study aimed to further our understanding of the trophic ecology of the Peruvian anchovy—a fish that supports the most productive fishery in the world. Existing data on anchovy diet, like that of other pelagic fishes, is largely derived from the analysis of stomach contents, and has generally concluded that anchovy production in the study region is fuelled by the consumption of phytoplankton (Medina et al., 2015) and copepods (Castillo et al., 2002; Castillo et al., 2011). Our results, based on stable isotope values, indicate that the contribution of phytoplankton to anchovy somatic tissues is minimal and that zooplankton represents the main food source assimilated by anchovies. Our results follow those of Espinoza and colleagues (Espinoza & Bertrand, 2008; Espinoza et al., 2017), who have elegantly demonstrated that a reliance on stomach content data has resulted in a significant misinterpretation of anchovy trophic ecology and by extension, the food web of the HCS. Beyond our reaffirmation of Espinoza et al.’s observations (2017) regarding the trophic position of anchovy, we have also identified several previously unknown features of the pelagic food web of the Chilean HCS, suggesting that the trophic ecology of the Peruvian anchovy, and ecosystem function in the HCS are both more complex than is typically considered.

The HCS has several characteristic features associated with elevated fisheries production. Using stable isotopes of anchovies, jack mackerel and their putative prey, as well as the key pelagic primary producer, phytoplankton, we were able to highlight the influence of upwelling throughout the food chain in the Chilean HCS. Phytoplankton and higher trophic levels were naturally-labelled with heavy nitrogen associated with upwelling (Casciotti, 2016; Docmac et al., 2017; Reddin et al., 2015), and results indicate that phytoplankton was the likely carbon source fuelling upper trophic levels in the HCS. We used phytoplankton collected across the survey area both as a putative food source in mixing models and as a baseline for the estimation of trophic position. Phytoplankton δ¹³C values varied by ca. 3‰ between locations, likely reflecting local differences in primary production rates due to variation in upwelling intensity, CO₂ concentration, phytoplankton growth rate etc., (Magozzi et al., 2017) and possibly community composition (Fry & Wainright, 1991). It is notable that several taxa examined here were notably ¹⁵N-enriched relative to their counterparts in the northern HCS as reported by Espinoza et al. (2017) (e.g., copepods by 7.5‰, euphausiids by 7.6‰, anchovies by 7.6‰ and jack mackerel by 4.1‰). This likely reflects regional differences in upwelling intensity.

Anchovies have typically been considered to have homogeneous diets and therefore trophic position (Medina et al., 2015). Using SIA, however, we have shown the existence of considerable inter-individual differences in δ¹⁵N, a robust indicator of trophic position (Post, 2002). The differences between the two approaches presumably reflects an overestimation on the contribution of phytoplankton due to using count data (Espinoza & Bertrand, 2008) as well as the different timescales associated with SCA (hours) compared with SIA (months) (Nielsen et al., 2018; Thomas & Crowther, 2015), but highlights the power of SIA to identify cryptic individual variation (Bolnick et al., 2002; Harrod et al.,  2005).

The observed variability in δ¹⁵N values for adult anchovies very likely reflects differences in individual feeding strategies. We identified two trophic groups, one relatively enriched in ¹⁵N, associated with the consumption of anchovy larvae, and another, depleted in ¹⁵N that largely consumed crustacean larvae. Anchovies from the first group had similar δ¹⁵N values to juvenile jack mackerel, implying that they fed at a similar TP: however they differed in δ¹³C values, likely indicating resource segregation. Anchovies in the lower δ¹⁵N group had a modal (95% credibility interval) estimated TP of 2.91 (2.62–3.23) indicating some omnivory, but a largely carnivorous diet. Anchovies in the higher δ¹⁵N group had a modal value of 3.79 (3.48–4.16) which overlapped completely with carnivorous jack mackerel (3.80 (3.51–4.14)), indicating that they are functionally similar. Support for a high TP for some anchovy is provided by Espinoza et al. (2017) who also showed overlap between jack mackerel δ¹⁵N values and anchovies with high δ¹⁵N values from southern Peruvian latitudes.

When the two anchovy groups were pooled, estimated TP (3.23 (2.93–3.58)) was ca. 1 TP higher than values estimated from stomach contents data and widely used in modelling studies (Guénette, Christensen & Pauly, 2008; Pauly et al., 1998). However, these pooled estimates effectively overlap with values from other studies using stable isotope analysis, such as Hückstadt, Rojas & Antezana (2007) working in central Chile (anchovy TP = 3.6) and Espinoza et al. (2017) from the Peruvian part of the HCS (3.4–3.7).

Our sample of anchovy was relatively small (n = 30), with individuals falling into two broad groups based on δ¹⁵N values. It is unknown whether the patterns in inter-individual variability in anchovy δ¹⁵N values we have shown are typical of the species, or even the genus. However, in their Fig. 5 Espinoza et al. (2017) show both a similar range of δ¹⁵N values (12–20‰), as well as possible evidence for different δ¹⁵N groups from anchovies from latitudes ca. 300–350 km north of where our samples were collected. It is possible that further intensive sampling may reveal that anchovy feed along a δ¹⁵N continuum rather than in discrete groups as we suggest. It is likely that this is worthy of further study in other populations and Engraulis species.

An alternative explanation for the existence of two groups of anchovy differing in δ¹⁵N (which we associate with differences in TP) is that the sample included fish from different geographical locations, where baseline δ¹⁵N differed, but that had actually fed at similar TPs. Baseline δ¹⁵N varies considerably along the Pacific coast of Chile and Peru, reflecting differences in the intensity of upwelling and the associated denitrification process. De Pol-Holz and colleagues (2009) reported a marked South–North gradient of ¹⁵N enrichment in marine sediments along the Chilean coast. We have shown elsewhere (Docmac et al., 2017) that consumer δ¹⁵N increases along a South-North gradient in the study region, and that δ¹³C co-varies spatially across the same scale. Given that the two putative groups of anchovy had similar δ¹³C values, we feel it is unlikely that they had different origins, and that we are confident in our identification of the two groups feeding at different trophic positions.

There was no evidence for any relationships between individual body mass and δ¹³C or δ¹⁵N in anchovy, as seen in the northern HCS by Espinoza et al. (2017), who examined fish from a larger size range. In our study, jack mackerel δ¹³C was similar across the size gradient, but showed a significant but small (0.6‰) positive shift in δ¹⁵N from the smallest (43.2 g) to the largest (79.8 g) individual examined. In contrast to our results, Espinoza et al. (2017), again in a wider size range, showed that jack mackerel δ¹⁵N was negatively related with individual size in the northern HCS. In order to have a reliable picture of ontogenetic shifts in δ¹³C and δ¹⁵N in N Chile, it is clear that larger samples are required both in terms of the number and of the size of individual fish.

Our estimates of anchovy consumption patterns and TP using stable isotope ratios are based on several key assumptions. For example, they are reliant on the use of representative values for the different prey groups and the isotopic baseline (phytoplankton) used to estimate TP, as well as for trophic discrimination (isotopic differences between prey and the consumer). With regard to our estimates of putative prey and the isotopic baseline, we followed a standardised sampling protocol across a considerable area (>2,000 km²), which is likely representative for the wider region. Such wide-scale sampling should be sufficient to account for spatial differences in phytoplankton and prey δ¹⁵N associated with variation in in upwelling intensity, and hence the relative enrichment in ¹⁵N associated with this process (Reddin et al., 2015).

However, upwelling intensity (and the relative amount of ¹⁵N-enrichment) in the region also varies on a temporal scale (Herrera & Escribano, 2006), raising the possibility of a disconnect between phytoplankton (which rapidly reflect isotopic changes) and anchovies, whose tissue δ¹⁵N values may reflect conditions in an earlier time period, e.g., when phytoplankton δ¹⁵N was different than during the study period. Such isotopic decoupling has been reported elsewhere (O’Reilly et al., 2002) and, if not identified, can potentially confuse estimates of fish TP. We do not think this is a major issue in our study as the presence of favourable winds year-round means than coastal upwelling occurs throughout the year in the study region (Palma, Escribano & Rosales, 2006; Thiel et al., 2007). This suggest that there is unlikely to be a marked seasonal shift in δ¹⁵N values at the base of the food web that could lead to us underestimating baseline δ¹⁵N. Furthermore, anchovy growth is rapid in the study region, and continues throughout the year (Cerna & Plaza, 2016), lowering the risk of seasonal decoupling from a dynamic isotopic baseline.

In terms of TDF, we used a commonly applied δ¹⁵N TDF of 3.4 ± 1.0‰ to allow direct comparisons with Espinoza et al. (2017). This value represents a mean estimated from many different habitats, feeding strategies and taxa (Post, 2002). Studies of carnivorous fish often use a smaller δ¹⁵N value of 2.9 ± 0.3‰, following McCutchan et al. (2003). Use of these values would increase our estimates of anchovy TP. Beyond the actual mean values used, it is also important to note that our estimates of TP were calculated following a robust Bayesian approach that includes error in both diet-tissue TDF and baseline δ¹⁵N (Quezada-Romegialli et al., 2018). Our approach however relies on the assumption that TDFs are additive, and this has been criticised by some authors. Caut, Angulo & Courchamp (2009) suggested that TDFs decrease with greater concentrations of ¹⁵N in the diet, and Hussey et al. (2014) developed a model to estimate what they refer to as scaled-TP. Their model includes a threshold (named δ¹⁵ N_lim) estimating the situation where δ¹⁵N in the diet is such that trophic discrimination is expected to be zero. δ¹⁵ N_lim is calculated based on estimates for the slope (β₀) and the intercept (β₁) of a negative relationship between consumer TDF and the δ¹⁵N of their food based on experimental feeding studies. Using the Bayesian median estimates provided by Hussey and colleagues based on a meta-analysis, δ¹⁵ N_lim has a value of 21.9‰. Many consumers in the HCS of N Chile including benthic rockfish (Docmac et al., 2017) and the pelagic fishes detailed here (jack mackerel and high δ¹⁵N anchovies) have δ¹⁵N values greater than this threshold. This not only prevents the calculation of scaled TP estimates in these naturally ¹⁵N-enriched consumers, but highlights that the relationship between experimentally-derived TDF and dietary δ¹⁵N at the heart of the scaled-TP approach does not extend to natural ecosystems where δ¹⁵N is naturally high. Although this does not discount the Hussey et al. (2014) method, it does highlight that it is not universally applicable, and raises the need for more experimental studies.

Our mixing model results reflect an attempt to estimate the relative contribution of seven different putative prey groups using only two different isotopes. Although Bayesian mixing models can provide useful results where the number of potential sources is greater than the number of markers used (Phillips et al., 2014), the performance of our model might have been limited by both the large number of prey groups and the fact that some of our prey groups were isotopically similar (See Fig. 2), even though these groups differed taxonomically (and functionally). We attempted to counter this by including information on prey elemental concentration in the model (i.e., we used a concentration-dependent model) which will likely be most useful in distinguishing between animal and phytoplankton prey (Phillips et al., 2014). Given these caveats, we feel confident that our mixing model results (Fig. 4), support our conclusions that 1. anchovy are not assimilating significant amounts of C and N from phytoplankton; 2. that the anchovy population includes individuals feeding on two broad groups of animal prey, i.e., consumption of crustacean larvae by the low δ¹⁵N anchovy group, and ichthyoplankton (eggs and preflexion-stage anchovy larvae) by the high δ¹⁵N anchovy group (as well as juvenile jack mackerel).

Our data indicate that anchovy diet is both more variable than is generally considered and that some individuals can feed at high trophic positions. It has become increasingly clear that populations of consumers can include individuals following specialised trophic strategies (Bolnick et al., 2003). The inclusion of different trophic specialisms within anchovy populations may contribute to their productivity by limiting intraspecific competition and support the unusually high fish production associated with the HCS (Bakun & Weeks, 2008; Chavez & Messié, 2009). The cannibalism and predation of eggs and larvae we suggest for the high δ¹⁵N anchovy group is commonly reported in clupeiform fishes (Alheit, 1987; Hunter, 1981), allowing consumption of an energy-rich prey (Konchina, 1991). The study area is recognised as a spawning and nursery area for fish, particularly anchovies (Palma, Escribano & Rosales, 2006; Palma, Pizarro & Flores, 1992). Given that anchovies and other fishes can spawn throughout the year in the region, this potentially results in a permanent supply of early life stage prey to predators, but may represent an important driver of natural mortality, with potential implications for future recruitment success.

Conclusions

We used a stable isotope approach to examine the trophic ecology of one of the world’s most important (but apparently understudied) fishes in an extremely productive, coastal upwelling zone. Our work has revealed several important findings with implications for how we understand and manage this important resource. We have shown that, like previous studies using stable isotopes from both central Chile (Hückstadt, Rojas & Antezana, 2007) and Peru (Espinoza et al., 2017), that anchovy TP has likely been considerably underestimated. Our estimates of anchovy TP differ from the classic value (2.2) used in much of the fisheries literature (Guénette, Christensen & Pauly, 2008; Pauly et al., 1998), and support Espinoza et al.’s (2017) observation that the classical hypothesis that short and efficient food chains drive secondary production in this ecosystem (Ryther, 1969) is no longer valid. We also have shown that the trophic ecology of individual anchovies is far more complicated than previously considered—individuals captured in the same net, fell into two broad trophic groups, even though they were largely of a similar size. These groups differed in trophic position and estimated diet, revealing previously unrecognised trophic variation in a fish that has long been considered to indiscriminately consume phytoplankton (i.e., TP∼2.2). Given the similarities with the data presented by Espinoza et al. (2017) from anchovies further north in their distribution (which also showed considerable variation in δ¹⁵N), it is clear that any model assuming that anchovy are feeding at a low TP is suspect, and management decisions based on such models should be reconsidered.

Our use of stable isotope analysis to characterise ecosystem function in the pelagic zone of the HCS has shown that the system is more complex than previously thought. This parallels a recent study where, using a similar approach, we challenged long-held assumptions regarding energy supply in coastal kelp forest ecosystems in the north of Chile (Docmac et al., 2017). The current study highlights the utility of stable isotope analysis as a means to reveal cryptic ecological variation in systems that are generally considered well described. We are currently examining amino acid δ¹⁵N of different pelagic fishes from northern Chile including anchovy and jack mackerel to examine whether we see similar estimates of TP (McClelland & Montoya, 2002) and levels of individual variation as seen here with isotope analysis of bulk materials.

Supplemental Information