Introduction

Coral reefs worldwide are at risk from geochemical and physical changes in the ocean driven by geologically unprecedented rates of CO2 release (Kiehl 2011; Hoenisch et al. 2012). Although the capacity of reefs to acclimate to rapid global change is uncertain, reef resiliency may be enhanced or protected by reducing local disturbance (Hughes et al. 2003; Hoegh-Guldberg et al. 2007). Understanding the trophic ecology of healthy, minimally impacted coral reef ecosystems improves the ability to evaluate resiliency and establish guidelines for protected areas (Hughes et al. 2003; Vroom and Braun 2010). Remote, minimally impacted coral reef ecosystems differ from disturbed coral reef ecosystems by their greater biomass and higher relative proportions of apex predators to herbivores (Friedlander and DeMartini 2002; Stevenson et al. 2007). The multiple species of herbivores, predators, and competitors of intact systems help prevent phase shifts in coral reef ecosystem trophic structure that lead to reductions in coral cover in favor of macroalgal cover and to reductions in reef resiliency to disease and bleaching (Sandin et al. 2008; Hixon 2011). Although often associated with reduced coral reef health in tropical systems (Hughes et al. 2010), high macroalgal cover is typical for subtropical systems (Harriott and Banks 2002). In relatively undisturbed subtropical systems, high macroalgal cover may provide critical habitat maintaining ecosystem health (Vroom and Braun 2010) or, in co-occurrence with low coral recruitment and low coral growth rates, limit potential resilience (Hoey et al. 2011).

The Northwestern Hawaiian Islands, designated as Papahānaumokuākea Marine National Monument (PMNM) in 2006, consists of atolls, reefs, islands, and submerged banks extending 2,000 km northwest of the main Hawaiian Islands (Friedlander et al. 2008). The PMNM is one of the few, relatively undisturbed, remote coral reef ecosystems serving as a reference point for restoration research (Knowlton and Jackson 2008). Like the unpopulated Kingman Reef and Palmyra Atoll in the Central Pacific Line Islands, the PMNM is dominated by apex predators (Friedlander and DeMartini 2002; Stevenson et al. 2007; Sandin et al. 2008). Unlike those tropical coral reef ecosystems, the subtropical PMNM has greater macroalgal cover than coral reef cover (Vroom et al. 2005). PMNM benthic communities are typified by dense, healthy patches of coral surrounded by hard-bottom habitat dominated by algae (Parrish and Boland 2004; Friedlander et al. 2008; Vroom and Braun 2010).

Understanding the trophic ecology of the PMNM includes determining the relative contribution of benthic primary production to the food web and the number of trophic positions between primary producers and apex predators. Stable isotope (SI) analyses have provided insight into energy pathways in coral reef ecosystem trophic ecology (Carassou et al. 2008; Greenwood et al. 2010), as has mass-balance ecosystem modeling. The ECOPATH mass-balance model was applied to the shallow region of French Frigate Shoals (FFS) in the PMNM to estimate mean annual biomass, production, and consumption for primary producers and consumers (Polovina 1984). Model input included assigning the contribution of benthic primary production to the atoll food web at 90 %, an assumption supported by comparisons of ECOPATH model results with metabolism studies (Atkinson and Grigg 1984). In contrast, an ECOPATH model coupling pelagic and benthic systems of lagoonal communities of an open atoll in New Caledonia predicted that benthic primary production was responsible for 50 % of total net primary production and 61 % of total consumption (Bozec et al. 2004). The FFS ECOPATH model predicted a food web of almost five trophic positions (TP) with piscivores, jacks, and sharks between positions four and five (Grigg et al. 2008). The New Caledonia model predicted just over four TPs with sharks and pelagic piscivores at the highest positions (Bozec et al. 2004).

Stable isotope analysis is commonly used to trace food sources and energy flow in food webs (Post 2002). Algal carbon (C) and nitrogen (N) SI composition is a function of ocean or sediment chemistry, photosynthesis, growth rates, and specific nitrogen uptake mechanisms. Typically, phytoplankton have lower δ 13C (~−22 ‰) than benthic algae (~−17 ‰) (France 1995), but similar δ 15N values (Owens 1987). Consumer SI values are higher than their food sources by a trophic enrichment factor (TEF) reported in mixing model literature as ranging from 0 to 1.4 ‰ Δ13C and 2.2 to 3.5 ‰ Δ15N (Galván et al. 2012). Because primary producer δ 13C signatures tend to be preserved in consumers at higher TPs, δ 13C values help distinguish the relative importance of different primary producers as sources of energy supporting the food web. Consumer δ15N values provide useful information about TP because of the relatively large TEF (Δ15N). Linear mixing models are based on mass-balance equations and the assumption that, after adjustment for TEF and TP, the consumer isotope ratios are identical to the isotopic base of the consumer’s food web (δ 15Nfood web base and/or δ 13Cfood web base), which is usually a mixture of potential primary producers. A range of feasible contributions by potential primary producer sources can be calculated and depend on the position of adjusted consumer isotope value(s) along a mixing line (one-isotope/two-source models) or within a polygon formed by values of three or more sources (primary producers) in a two-isotope mixing model. Mixing models and the relationships between TEF, TP, and consumer values were used to evaluate trophic relationships based on a large dataset of consumer and primary producer SI values collected, opportunistically, over the length of the PMNM in 3 years (2001, 2004, and 2005).

Methods

Sample collection and shipboard processing

Primary producer (n = 50) and consumer (n = 549) samples were collected for SI analysis between August 2001 and October 2005 on four research cruises at locations spanning the length of the PMNM (Fig. 1). Sampling was temporally and spatially opportunistic, and the majority of samples were obtained from six locations: Kure Atoll, Midway Island, Pearl and Hermes Atoll, Maro Reef, FFS, and Necker Island (Fig. 2). Targeted species include those classified in trophic groups as herbivore and apex predator by Friedlander and DeMartini (2002); as herbivore, zooplanktivore, benthic carnivore, and piscivore by Parrish and Boland (2004); and the most common macroalgae Microdictyon and Halimeda (Parrish and Boland 2004). Primary producers included three main groups: phytoplankton, benthic microalgae (BMI), and benthic macroalgae (BMA) (Table 1). In 2005, we collected 314 samples on NOAA ship Hi‘ialakai cruises HI-05-04 (May) and 253 samples during the HI-05-10 Northwestern Hawaiian Islands Reef Assessment and Monitoring Program (NWHIRAMP) (September–October). An additional 32 samples from cruises in August 2001 and August 2004 were obtained to double the number of macroalgae SI measurements and provide samples from otherwise unsampled locations (Fig. 2). With the exception of eight Caranx samples collected in September 2004 from FFS and Necker Island, all consumers were sampled in 2005 (Fig. 2; Table S1). Benthic macroalgae and shark sampling mostly occurred at depths >30 m; invertebrates were collected in depths <20 m; and fish and BMI were sampled in both shallow and deeper waters (3–60 m). Seawater (6–10 L) was collected from four sites with Niskin bottles from 1 to 35 m depths. Water samples were pre-filtered through 200-μm mesh to remove zooplankton and then filtered on pre-combusted glass fiber filters (GF/F) to obtain phytoplankton samples. Surface (top 0.5–1 cm) sediment samples were collected by divers or by Ponar grab, and BMI were separated using two methods. Eighteen samples from five sites were concentrated on pre-combusted glass fiber filters (GF/F) following vertical migration (VM) through a 63-μm Nitex mesh and silica layer (Wainright et al. 2000). In addition, six samples from two locations were collected on pre-combusted glass fiber filters (AH) following rinsing and density centrifugation with Ludox (Lu; colloidal Si) (Moseman et al. 2004). Both techniques were utilized in an attempt to sample the entire BMI community. All filtered phytoplankton and BMI samples were frozen immediately. In May 2006, they were thawed, fumed with concentrated HCl to remove carbonates, and shipped for SI analysis.

Fig. 1
figure 1

Map of the Hawaiian Archipelago showing the locations sampled in the Papahānaumokuākea Marine National Monument. Map modified from Marine Mammal Commission (2001)

Full size image
Fig. 2
figure 2

Spatial and temporal sampling frequency of major primary producer (a) and consumer groups (b–e). Locations are ordered geographically from northwest to southeast. Location abbreviations are Kure (Kure Atoll), Mid (Midway Island), P&H (Pearl and Hermes Atoll), Nor (Northampton Seamounts), Lay (Laysan Island), Maro (Maro Reef), Raita (Raita Bank), Gard (Gardner Pinnacles), FFS (French Frigate Shoals), Necker (Necker Island), MB (Middle Bank), and unknown. In the first panel, each location is bordered by six gray vertical lines representing sampling months in chronological order. Lines representing August 2001, August 2004, and September 2004 precede each tick line, and lines representing May 2005, September 2005, and October 2005 follow each tick line. Lines are extended into panels (b–e) for months in which consumers were sampled. Note the change in the vertical scale between primary producer and consumer panels

Full size image
Table 1 Stable isotope values of the primary producer groups, combined groups, subgroups, and species
Full size table

Fifty samples of 13 BMA genera (Table 1) (~100 mg wet weight, above-ground portion) were collected by divers. Benthic macroalgae samples were rinsed in freshwater and frozen immediately. All BMA samples were thawed in January 2008, rinsed with distilled water and cleaned of epiphytes, photographed under a dissecting microscope to aid identification, dried at 60 °C, ground with a mortar and pestle or Wig-L-Bug® grinding mill, acidified with 1 N HCl to remove carbonates, rinsed, and redried in preparation for SI analysis. Some samples were also acidified prior to grinding.

Invertebrates were captured by divers, and fish and sharks were captured by divers or by hook and line. Larger specimens, including sharks and jacks, were released live after white muscle plugs were obtained. White muscle tissue from smaller specimens was obtained by dissection. Samples were frozen immediately after collection and stored until processing (2006–2008). In the laboratory, muscle tissue was thawed, rinsed with distilled water, dried at 60 °C, and ground with a mortar and pestle or Wig-L-Bug® grinding mill prior to SI analysis. We did not correct δ 13C values for lipid content, as all consumer groups had C:N ratios of 3.4 or less (Post et al. 2007).

Stable isotope analysis

Dried, ground samples were weighed, placed into tin capsules, and sent to the University of California-Davis (UC-D) SI Laboratory (n = 588) or the University of Washington (UW) SI Core Laboratory (n = 11) for analysis. Replicate samples were run on a regular basis to check for consistency, and a few of the reported results represent the average of two replicates. Stable isotope analyses were conducted using a Sercon Ltd. PDZ Europa ANCA-GSL elemental analyzer interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (IRMS) (UC-D) or a continuous flow Thermo Finnigan Delta Plus XP IRMS and Costech Analytical ECS 4010 elemental analyzer (UW). Carbon and nitrogen isotope ratios are reported in units per mil (‰) relative to the standards Vienna Peedee Belemnite (VPDB) for carbon and atmospheric nitrogen (air) for nitrogen:

$$\delta^{13} {\text{C}} = \left[ {\left( {\frac{{Ratio_{sample} }}{{Ratio_{VPDB} }}} \right) - 1} \right] \times 1,000 $$
(1)
$$\delta^{15} {\text{N}} = \left[ {\left( {\frac{{Ratio_{sample} }}{{Ratio_{Air} }}} \right) - 1} \right] \times 1,000 $$
(2)

Trophic group assignment

Consumers were classified into trophic groups. For species not classified by Parrish and Boland (2004) or Friedlander and DeMartini (2002), we referred to Hiatt and Strasburg (1960), Hobson (1974), Allen et al. (1998), DeFelice and Parrish (2003), and Piché et al. (2010) for trophic information (Table 2). We classified six species as herbivores (n = 119), four species as zooplanktivores (n = 77), and the butterflyfish Chaetodon lunulatus (n = 15) was the sole corallivore. Piché et al. (2010) included invertebrates and vertebrates in their benthic carnivore group, but we created two classes of benthic carnivores. The benthic carnivore (invertebrate) group included a single Octopus cyanea sample and 42 lobsters (Panulirus marginatus). Seven fish species were classified as benthic carnivore (vertebrate) (n = 138) including the introduced snapper Lutjanus kasmira. L. kasmira was classified as an omnivore by Piché et al. (2010) based on diet information in FishBase (Froese and Pauly 2012), whereas we based our classification on the DeFelice and Parrish (2003) report. The piscivore group (n = 21) consists of two species that occupy different habitats: the benthic bigeye (Priacanthus meeki) and the pelagic tuna (Euthynnus affinis). An alternate classification for P. meeki is zooplanktivore (Piché et al. 2010). Seven species (a snapper, two jacks, a grouper, and three sharks) were classified as apex predators (n = 104). Trophic position assignments (Table 2) were based on estimates of the ECOPATH model of FFS [modified from Polovina (1984) in Grigg et al. (2008)] with a few exceptions. The corallivore group (unrepresented in the ECOPATH model) was assigned the FishBase TP of 3.3 (Froese and Pauly 2012). ECOPATH estimated cephalopods at TP 3.9, but we included our single specimen with lobsters in the benthic carnivore (invertebrate) group (TP 3.5). ECOPATH estimates for reef sharks (TP 4.7) were higher than their estimates for G. cuvier (TP 4.5) and jacks (TP 4.1), but we placed all apex predators excluding G. cuvier at TP 4.0, consistent with most FishBase estimates (Froese and Pauly 2012).

Table 2 Stable isotope values for consumer groups and species
Full size table

Statistics and modeling

Ontogenetic shifts in shark diet are associated with species-specific size classes (Lowe et al. 1996; Wetherbee et al. 1996, 1997). We examined the relationship between total length and δ 15N for six Carcharhinus amblyrhynchos (medium and large size classes), 16 Carcharhinus galapagensis (small, medium, and large size classes), and eight Galeocerdo cuvier (large size class) that were sampled in May and September 2005 at four locations. Fork length (FL), recorded instead of total length (TL) for two C. amblyrhynchos and one C. galapagensis specimens, was converted to total length using ratios (1.2 TL:FL) measured during this study.

Although sampling frequency, distribution, and replication were inadequate for spatial, temporal, or comparative analysis of the major groups, subgroups, and species (Fig. 2), SI values averaged by subgroup, species, location, and date (Table S1) provide clues to potential sources of the observed variability. We used mixing models to analyze trophic relationships between consumer and primary producer groups. Mixing model solutions depend on the isotopic values of primary producer source groups and values of TEF and TP used to adjust consumer values to the base of their food web. The relationship between TEF, TP, consumer isotopic values, and the isotopic value of the base of the food web is defined by Eq. (3).

$${\text{X}}_{food\,web\,base} = {\text{X}}_{consumer} - [(TP - 1) \times TEF] $$
(3)

where X is either δ 13C or δ 15N.

We first estimated the percent contribution of phytoplankton and benthic (combined BMI and BMA) primary production to consumer production using the one-isotope, two-source mixing model, IsoError (Phillips et al. 2005). Model input included the δ 13C mean ± standard deviation (SD) of both primary producers and each consumer group (after adjustment to the base of the consumer’s food web). δ 13Cfood web base values were calculated using Eq. 3, an assumed TEF of 0.5 Δ13C (McCutchan et al. 2003), and the assigned TP (Table 2). Results are reported as mean and 95 % confidence intervals (CI) of percent contribution to consumer production.

We also attempted to estimate the separate contribution by BMI and BMA and utilize δ 15N data with a three-source, two-isotope mixing model. A requirement of three-source mixing models is that adjusted consumer values lie within the polygon formed by the SI values of sources (primary producer groups). We found that the low and similar  δ 15N means of the three main groups (phytoplankton, BMI, and BMA) resulted in a source polygon too small to solve the model unless consumers were adjusted using Δ15N (TEF) or TP values that were higher than literature values. Therefore, we used Eq. 3 with a range of literature values for Δ15N TEF and the TPs assigned in Table 2 to explore feasible values for δ 15Nfood web base. Two alternate (solvable) mixing model source polygons that included sources with higher δ 15N values were created by (1) increasing the δ 15N value of the phytoplankton group above the measured mean to a value consistent with literature values (Owens 1987) and (2) replacing the BMI and BMA endmember groups with four subgroups of benthic algal values (High N BMI and BMA, Low N BMI, and Low N BMA).

Results

Primary producers

Primary producer SI values averaged −16.2 ‰ δ 13C and 2.1 ‰ δ 15N (Table 1). Mean δ 15N values of the three main groups were 1.4 (phytoplankton), 1.5 (BMA), and 3.0 ‰ (BMI) (Table 1). Within groups, the range of δ 15N values was ≥5.5 ‰ (Table 1; Fig. 3). Relative to BMI and BMA, phytoplankton had a narrow range of δ 13C values and a low δ 13C mean. The eight BMI (Lu) measurements had a smaller range in SI values than 18 BMI (VM) measurements (Table 1; Fig. 3). Six of the BMI (VM) samples (from three locations) had δ 15N values >4.5 ‰ and relatively low values of δ 13C (<−12.4 ‰) (Fig. 3). The δ 13C and δ15N values of these six measurements (the subgroup High N BMI) averaged 8.0 ‰ lower and 3.6 ‰ higher, respectively, than the average of the remaining 20 measurements (Low N BMI) (Table 1). Benthic macroalgae also exhibited a wide range of SI values (Table 1; Fig. 3). Five samples (two unidentified red algae, two Sporochnus dotyi, and one Portieria hornemanni) from three locations had δ 13C values <−27 ‰ (Fig. 3) and formed a Low C Low N BMA subgroup (Table 1). With the exception of these five specimens, BMA δ 13C values were ≥−21.8 ‰ (High C Low N BMA subgroup; Table 1). Four δ 15N values between 5.3 and 8.0 ‰ were obtained in September 2005 at Kure Atoll (Dictoyta friabilis and Padina sp.) and in October 2005 at Necker Island (Halimeda velasquezii and Caulaerpa racemosa) (Table S1). The High N BMA subgroup formed by these four specimens had δ 13C and δ15N means similar to those of the High N BMI subgroup, and the δ 15N values of the remaining BMA (Low N BMA subgroup) ranged from −1.6 to 3.5 ‰ (Table 1). Although the Low C Low N BMA and High N BMA subgroups had δ 13C and δ 15N values that were 11.4 ‰ lower and 5.8 ‰ higher (respectively) than the remaining 41 measurements (the High C Low N BMA subgroup), they had minimal influence on the overall BMA mean (Table 1). Finally, the averages of all (combined) BMI and BMA δ13C and δ15N values were slightly higher than BMA means (Table 1).

Fig. 3
figure 3

δ 13C (a–c) and δ 15N (d–e) values of individual phytoplankton, benthic microalgae (BMI), and benthic macroalgae (BMA) samples are presented in spatial and temporal context. The relative position of vertical gray lines represents months sampled at each location in chronological order. Lines representing August 2001, August 2004, and September 2004 precede tick lines, and lines representing May 2005, September 2005, and October 2005 follow tick lines. Only months sampled for primary producers are shown; refer to Fig. 2a for relative spacing of lines representing all 6 months. Benthic microalgae are separated by methodology: VM vertical migration and Lu Ludox. Benthic macroalgae values are shown by phyletic groups. Samples with δ 15N values >4.5 ‰ are indicated by the larger circles. See Fig. 2 for location abbreviations and Table S1 for average values per location and month. VPDB Vienna Pee Dee Belemnite

Full size image

Consumers

Consumer group δ 13C means were centered near the combined BMI and BMA mean (−15.6 ‰) although zooplanktivore (−16.9 ‰) and piscivore (−17.8 ‰) means were slightly lower and the corallivore mean (−13.2 ‰) was slightly higher (Table 2). Herbivores had a wider range of individual δ 13C values than all other consumer groups (Table 2; Fig. 4) and a δ 15N mean (6.6  ‰) that was 4.5 ‰> the mean of all sampled primary producers (Table 1). The zooplanktivore, corallivore, and the benthic carnivore (invertebrate) groups had similar δ 15N mean values (~8.3 ‰) as did the vertebrate benthic carnivore and piscivore groups (~9.3 ‰) (Table 2). Apex predator δ 15N mean values were higher than piscivore and benthic carnivore (vertebrate) groups by ~1 ‰ (excluding Galeocerdo cuvier) to ~3 ‰ (G. cuvier) (Table 2).

Fig. 4
figure 4

The individual carbon and nitrogen stable isotope values of each consumer species plotted by trophic group. Apex predators are shown in two panels, G (jacks and jobfish) and H (sharks). The full names and basic statistics for each species are provided in Table 2. VPDB Vienna Pee Dee Belemnite

Full size image

Within consumer groups, several species exhibit distinctive isotopic signatures, including some with apparent spatial variability. Relative to all other species, the herbivore Zebrasoma flavescens stands out as having the lowest δ 13C and δ 15N means and largest range of δ 13C values (Table 2; Fig. 4). The herbivore Ctenochaetus cf. strigosus (n = 4) and the sole corallivore Chaetodon lunulatus had similar δ 13C means (~−13.3 ‰) that were at least 0.9 ‰ higher than means of all other consumer species (Table 2). The few individual herbivore δ 15N values >7.5 ‰ were found only at Kure Atoll (three species) and Necker Island (one species) (Table S2). The wide range in zooplanktivore δ13C values was driven by Chaetodon miliaris, the only zooplanktivore with δ13C values >−16.3 ‰. The eight highest C. miliaris δ13C values (>−14.5 ‰) occurred at Kure Atoll (Table S2), and relatively high δ13C values were also found at Kure Atoll for the lobster Panulirus marginatus (Table S2). Individual zooplanktivore δ 15N were, generally, ~1 ‰ lower at FFS and Necker Island than at other locations (Table S1).

Three pairs of benthic carnivore (vertebrate) species had similar SI signatures or patterns in variability (within pairs). Chaetodon frembii and Lutjanus kasmira had similar δ13C and δ 15N means. Bodianus bilunulatus and Parupeneus porphyreus had similar δ13C ranges and means, and, for both species, the highest individual δ13C values (>−12 ‰) occurred at Midway Island in May 2005 (Table S2). Although Parupeneus multifasciatus mean SI values were lower than those of Thalassoma ballieu (Table 2), both species exhibited similar spatial variability with lower δ13C means and ranges at Maro Reef and higher δ 15N means and ranges at Necker Island (Table S1). In addition, the T. ballieu δ 15N mean (10.2 ‰) was highest of all benthic carnivores and equal to the mean of apex predators (excluding Galeocerdo cuvier) (Table 2).

Euthynnus affinis had the second lowest δ13C mean of all consumer species and differed from the only other piscivore Priacanthus meeki by a wider δ13C range and higher δ 15N mean (Table 2). Except for Hyporthodus quernus (n = 2) and Aprion virescens (n = 6), all apex predator species exhibited ranges in δ13C values ≥2.6 ‰ (Table 2), and Caranx ignobilis had the widest range in δ13C values (6.5 ‰). Average δ15N values of A. virescens were similar to piscivore and benthic carnivore (vertebrate) means (~9.3 ‰). G. cuvier specimens were at least 90 cm longer than other apex predators and had average δ15N values 1.6–2.7 ‰> the means of other apex predator species. There was a positive but insignificant relationship between δ 15N and TL for G. cuvier (n = 8, from two locations) and Carcharhinus amblyrhnochos (n = 6, all from FFS) (Fig. 5). δ 15N values of large, medium, and small Carcharhinus galapagensis (n = 16, from four locations) overlapped with those of medium and large C. amblyrhynchos.

Fig. 5
figure 5

The relationship between total length and δ 15N values for three shark species. Individuals are plotted by species-specific size classes (Lowe et al. 1996; Wetherbee et al. 1996, 1997) and regression lines are shown for each species

Full size image

Mixing models

Results of the one-isotope (δ 13C), two-source mixing model, IsoError (Phillips et al. 2005) indicated that benthic primary production contributed, on average, a minimum of 65 % to consumer production with a lower minimum contribution (40 and 56 %, respectively) to piscivore and zooplanktivore groups than to other groups (≥66 %) (Table 3). Except for the corallivore group, mean benthic primary production contributions ranged from 53 (piscivore) to >90 % (both apex predator groups). The adjusted corallivore δ 13C value was higher than the means of both primary producer sources, which resulted in a negative phytoplankton contribution and a benthic primary producer contribution >100 %. The selection of TEF affected model solutions. Increasing Δ13C had the effect of lowering adjusted consumer values (δ 13Cfood web base) closer to the phytoplankton mean and away from the benthic primary producer mean. For each incremental 0.5 ‰ increase in TEF, the average relative contribution by benthic primary production decreased by 7–22 % with largest changes for consumers at higher trophic levels (not shown). Decreasing Δ13C had the opposite effect.

Table 3 Results of the two-source [phytoplankton and combined benthic microalgae (BMI) and benthic macroalgae (BMA)], one-isotope (δ13C) mixing model, IsoError
Full size table

Although consumer and primary producer individual δ15N values had similar ranges (9.2 and 9.6 ‰, respectively), the range in the δ15N means of three main primary producer groups was compressed (1.6 ‰) relative to the range in consumer group means (5.5 ‰) (Tables 1, 2; Fig. 6). This disparity imposed tight modeling constraints on the selection of TEF and TP values used to adjust consumer means to the base of the food web. Assuming the ECOPATH TP of 2.2 (Grigg et al. 2008) and applying the full range of literature values for Δ13C and Δ15N TEFs (Galván et al. 2012), all of the calculated herbivore δ 15Nfood web base values were outside of the mixing model polygon formed by the three major primary producer groups (Fig 7a). This suggests that either TP or Δ15N was too low or that an important source of primary production with higher δ 15N values was underrepresented. There is no evidence that the presumed herbivores have a higher TP than the ECOPATH estimate of 2.2. Although the 4.5 ‰ difference between the δ15N means of the herbivore (Table 2) and combined primary producers (Table 1) suggests a TEF of 3.8 ‰ Δ15N (calculated as using Eq. 3 and the TP of 2.2), this is 0.3–1.6 ‰ higher than all published estimates (Galván et al. 2012). We note also that this estimate of TEF is based on the assumption that the sampled primary producers represent the base of the herbivore food web. Another estimate of TEF is the difference between consumers and their presumptive prey, which averages 1.8 ‰ Δ15N for the sampled consumer groups. With no compelling support for an herbivore TP >2.2 or that PMNM consumers have a TEF > literature values, it is likely that primary producers with δ15N values higher than the means of the sampled primary producers were important to the PMNM food web. This conclusion is supported by a number of published primary producer δ15N values [e.g., Owens (1987) and (MacArthur et al. 2011)].

Fig. 6
figure 6

Primary producer and consumer group δ 13C and δ 15N means (± SD; large symbols) plotted in relationship to individual primary producer values (small symbols). The ranges of consumer and primary producer values are represented by boxes. BMI benthic microalgae, BMA benthic macroalgae, G. cuvier Galeocerdo cuvier, VPDB Vienna Pee Dee Belemnite

Full size image
Fig. 7
figure 7

Three mixing model source polygons are shown relative to δ 15Nfood web base and δ 13Cfood web base values (open circles) of the herbivores a and all consumer groups b and c. Isotopic signatures of the base of food web were calculated using Eq. 3, assumed TPs (Table 2), and TEFs of 0.5 ‰ Δ13C and 2.5 ‰ Δ15N. The gray box a represents the full range of potential herbivore food web base values calculated using a TP of 2.2 and literature ranges of TEFs (0–1.4 ‰ Δ13C and 2.2–3.5 ‰ Δ15N). The polygon formed by the average values of three main primary producer groups is shown in a. Alternate polygons were formed by increasing the measured phytoplankton average δ 15N value (1.4 ‰) to 4.0 ‰ b and by forming subgroups of benthic algae (Table 2) c. Benthic microalgae and benthic macroalgae are abbreviated as BMI and BMA, respectively. Error bars represent standard deviation. VPDB Vienna Pee Dee Belemnite

Full size image

We used Eq. 3 and a range of TEFs (2.0–3.5 ‰ Δ15N) (Galván et al. 2012) and δ 15Nfood web base values (2.0, 3.0, and 4.0 ‰ δ 15N) to calculate potential TPs for herbivore and Galeocerdo cuvier groups. The calculated TPs were compared with ECOPATH TPs to evaluate feasible TEF and δ 15Nfood web base values (Fig. 8). Calculated TP varied inversely with TEF and δ 15Nfood web base, and the sensitivity of TP increased with higher values of either variable. For δ 15Nfood web base values of 2.0, 3.0, and 4.0 ‰, the ECOPATH TP for G. cuvier could be calculated using TEFs ranging of 2.9, 2.6, and 2.3 ‰ Δ15N, respectively. The herbivore ECOPATH TP could be calculated for δ 15Nfood web base values of 3.0 and 4.0 using TEFs of 3.0 and 2.2 ‰ Δ15N, respectively. However, the herbivore ECOPATH TP could not be calculated for a δ 15Nfood web base of 2.0 ‰ unless a TEF > literature values was applied. This exercise demonstrates that a TEF near 2.5 ‰ Δ15N is feasible and that a primary food source with an average δ 15N value of 3 ‰ or greater was needed to solve a two-isotope, multi-source mixing model for the herbivore group using the ECOPATH TP, as suggested by Fig. 7a.

Fig. 8
figure 8

Calculated potential trophic positions (TP) for the herbivore group (solid black line) and Galeocerdo cuvier (dashed black line) are shown relative to ECOPATH TP estimates (horizontal lines) for herbivores (2.2) and G. cuvier (4.5). Potential TP was calculated using Eq. 3 and a range of TEFs (2.0, 2.5, 3.0, and 3.5 ‰ Δ15N) for three different δ 15Nfood web base values (2.0, 3.0, and 4.0 ‰). Arrows indicate the TEF value required to calculate ECOPATH TPs for each consumer using the different δ 15Nfood web base values

Full size image

To explore the isotopic signature of potentially important missing or undersampled sources of primary production, we created two alternate source polygons. The first was formed by increasing the average phytoplankton δ15N value from 1.4 to 4.0 ‰ (Owens 1987) (Fig. 7b). For the second, we created a subgroup of the six BMI and four BMA samples with δ15N values >4.5 ‰ and with similar ranges of δ13C values (High N BMI and BMA) (Table 2; Fig. 7c). The remaining BMI and BMA samples formed the Low N BMI and Low N BMA groups. Both polygons were sufficient to solve mixing models for all consumer groups which were adjusted to their food web base values using TPs listed in Table 2 and TEFs of 0.5 ‰Δ13C and 2.5 ‰Δ15N. Within each polygon, adjusted consumer δ15N values formed two clusters, with values of zooplanktivore, herbivore, and G. cuvier somewhat higher than other consumers (Fig. 7b, c).

Discussion

Primary producers

We observed large ranges in δ13C and δ15N values of primary producers with no strong evidence of phyletic, temporal, spatial, or methodological trends (Fig. 3). Much of the variability in BMI and BMA was driven by a few samples (BMA δ13C values ≤ −27.6 ‰ and BMI and BMA δ15N values >4.5 ‰) that had little effect on the mean values of the two groups (Table 2). The δ 13C results were typical for marine phytoplankton and BMA (France 1995; MacArthur et al. 2011); however, the BMI δ 13C results were ~8 ‰ lower than the average reported in France (1995) and may be indicative of cyanobacteria communities or a higher ratio of photosynthetic HCO3 −:CO2 utilization (Yamamuro et al. 1995; Kolasinski et al. 2011). Coral, unsampled in this study, is also a source of benthic primary production, but it is unlikely that its inclusion would have altered our results, as PMNM benthic algae δ13C values encompass the range of reported coral flesh and zoozanthellae δ13C values from around the world (−15.1 to −10.4 ‰) (Heikoop et al. 2000).

Based on the mean δ 15N values (1.4–3.0 ‰) of the three main groups of primary producers, N2 fixation appears to be an important source of N to both planktonic and benthic algal communities. However, an undersampled primary producer source group with δ 15N values >3 ‰ was required to solve a two-isotope, multi-source mixing model. Consumers may have preferentially utilized the High N BMI and BMA subgroup or phytoplankton with higher δ 15N values. However, we note that only one individual (Halimeda velasquezii) in the High N BMA group was a member of the genera representing the majority of PMNM primary production biomass. In addition, several other Halimeda specimens had lower δ 15N values (Table S2). Alternatively, phytoplankton and/or benthic primary producers with higher δ 15N values may have been more prevalent on an annual basis than in our May–October sampling window.

Seasonality, or a temporal lag between primary producer and consumer values, is often the source of putative variability in TEF (Hannides et al. 2009; Wyatt et al. 2010). Integration of the isotopic signature of food sources in consumer muscle tissue can take several months in some species (Logan and Lutcavage 2010; Madigan et al. 2012), and variability in primary producer δ 15N values can be driven by seasonality in the N sources for new production in the water column (Dore et al. 2002; Hannides et al. 2009) and the isotopic signature of N in advected or upwelled waters (O’Reilly et al. 2002).

Primary producers with lower δ 15N values may also be important to the PMNM food web, as indicated by the position of the adjusted values of several consumer groups within the alternate polygons (Fig. 7a, b). There is little fractionation associated with N2 fixation (Fogel and Cifuentes 1993), and low δ 15N values (~2.0 ‰) are consistent with a primary producer community utilizing N originating with N2 fixers (Dore et al. 2002). Particularly low near-reef macroalgae δ 15N values (~0.3 ‰) have been suggested to originate from cyanophytes (France et al. 1998), and even lower values are reported for Trichodesmium in cultures and field studies (McClelland et al. 2003). The importance of benthic cyanobacteria in food webs has been demonstrated in a diet study in Mariana Islands coral reefs (Cruz-Rivera and Paul 2006) and an SI study of a California salt marsh (Currin et al. 2011), while Trichodesmium values were the source of low zooplankton δ 15N values found in the tropical North Atlantic (McClelland et al. 2003).

Consumers

IsoError mixing model results indicated that benthic primary production (potentially including unsampled coral) provided the majority of support for the PMNM consumer groups (Table 3). However, consumer groups encompass species with a range of habitats and feeding habits, so it is worth noting instances where the SI values of species or individuals differ from their assigned trophic group. For example, similar δ13C values of the corallivore Chaetodon lunulatus; the herbivore Ctenochaetus cf. strigosus; and some individuals of the zooplanktivore Chaetodon miliaris (Fig. 4) suggest mutual dependence on sources with high δ13C values, which could include BMI or unsampled coral. The herbivore Zebrasoma flavescens had a δ13C mean similar to that of the piscivore group (the midwater tuna Euthynnus affinis and the benthic bigeye Priacanthus meeki). The wide (similar) ranges in Z. flavescens and E. affinis δ13C values (~−22 to −16 ‰) (Table 2) suggest these are opportunistic or generalist feeders (Bearhop et al. 2004) with some individuals perhaps more dependent on a phytoplankton-based food web (or, alternatively, a food web based on BMA species with relatively lower δ13C values).

Similar δ13C values of P. meeki and the Dascyllus and soldierfish zooplanktivores suggest a common reliance on primary producers with lower δ13C values. The relatively higher δ15N values of P. meeki support its inclusion in the piscivore group (Table 2). Also based on relative δ 15N values, the benthic carnivore Thalassoma ballieu may have been feeding at a higher TP than other benthic carnivores, and the apex predator Aprion virescens may have been feeding at a TP similar to that of the piscivores and benthic carnivore (vertebrate) groups (Table 2). The remarkable similarity of Chaetodon fremblii and Lutjanus kasmira δ 13C and δ 15N values suggests that these species occupy similar trophic niches; both were classified as benthic carnivores, but an alternate classification is omnivore (Piché et al. 2010; Froese and Pauly 2012). The wide range in Caranx ignobilis δ13C and δ 15N values may be indicative of an opportunistic diet (Bearhop et al. 2004) or individual dietary specialization (Matich et al. 2011).

The lack of significant correlation between length and δ 15N values for PMNM shark species and the overlapping δ 15N values of species-specific size classes provided no SI evidence for ontogenetic changes in diet. However, overlapping δ 15N values of large Carcharhinus amblyrhynchos and small Carcharhinus galapagensis are consistent with the high dietary overlap observed in a previous study (Papastamatiou et al. 2006). With a δ 15N mean value ~2 ‰ > other apex predators, Galeocerdo cuvier, at least 90 cm longer than other apex predators, appears to have been feeding at a higher TP. There was no isotopic support for the FFS ECOPATH model estimates placing the reef shark C. amblyrhynchos at a higher TP than G. cuvier.

There were indications of apparent spatial variability in some primary producer and consumers. Higher δ13C values of C. miliaris and P. marginatus occurred at Kure Atoll, and, for some benthic carnivore species, lower δ13C values were observed at Midway Island and Maro Reef. Within trophic groups, most exceptionally high δ 15N values were from Kure Atoll and Necker Island (Table S2). These include BMA values >5 ‰ δ 15N (n = 4), herbivore values >8 ‰ δ 15N (n = 4), benthic carnivore values >10.8 ‰ δ 15N (n = 7), and the G. cuvier value of 13.7 ‰ δ 15N (n = 1) (Fig. 3; Table S2). Although the trend was not observed in all trophic groups (e.g., zooplanktivore δ 15N values were low at Necker Island and FFS relative to other locations), some of the locational differences are consistent with previous findings. Significant differences in δ 15N values of the spiny lobster Panulirus marginatus at two PMNM locations were traced to relatively higher δ 15N values at the base of the Necker Island food web (O’Malley et al. 2012).

The data highlight the need for identifying sources of apparent spatial isotopic variability in consumers and for investigating the temporal scale of external and internal nitrogen supply processes and phytoplankton dynamics that influence primary producer isotopic variability. Observed spatial variation in SI values of New Caledonia sources and consumers have been ascribed to heterogenetic water column productivity (Carassou et al. 2008), and higher δ 13C and lower δ 15N values are associated with Trichodesmium (Carpenter et al. 1997). In addition, diel, tidal, lunar, and seasonal changes in connectivity among PMNM apex predator species (Dale et al. 2011) can complicate interpretations of spatial variability between sites. Variability in the observed consumer SI values may also be driven by ontogenetic variations, omnivory, or tissue turnover time (Sweeting et al. 2005).

Mixing models

Mixing model evidence for a major contribution of benthic primary production to the PMNM food web is consistent with ecosystem mass-balance model results. The average contribution by benthic primary production to the piscivore group was slightly lower than the 61 % benthic algal contribution for the entire food web in the ECOPATH model of a New Caledonia lagoon (Bozec et al. 2004). For other consumer groups, the average benthic primary producer contribution was 70 % or greater (Table 3), but, generally, lower than the 90 % assigned for the FFS food web in the FFS ECOPATH model (Polovina 1984). Mixing model performance is highly sensitive to TEF (Bond and Diamond 2011), and qualitative differences in model solutions can be examined using a range of TEF values (Galván et al. 2012). If the isotopic signatures of utilized primary producers were adequately represented in mixing models, the SI analysis would provide a test of TP estimates and TEF assumptions. However, this study provided qualitative evidence that the base of the PMNM food web included a source with a higher δ 15N value than the averages of the sampled primary producer groups. For alternate mixing model polygons (including a high δ 15N source), model solutions required a TEF in the lower range of literature values (2.5 ‰ Δ15N) (Fig. 7b, c). A low TEF (1.8 ‰ Δ15N) was also supported by differences between consumer δ 15N values. A TEF of 3.4 ‰ Δ15N (DeNiro and Epstein 1981; Vander Zanden and Rasmussen 2001; Post 2002) is frequently cited, but evidence for lower TEF values (McCutchan et al. 2003; Vanderklift and Ponsard 2003; Caut et al. 2009) is increasing. In a paired tissue and gut sample analysis of 152 coral reef fish from Western Australia, Wyatt et al. (2010) found an average TEF of 2.4 ‰ Δ15N. In addition, a TEF of ~2 ‰ Δ15N was required to match the North Pacific subtropical gyre zooplankton TP calculated using compound-specific isotope analysis (Hannides et al. 2009), and a TEF of 2.5 was required to match diet-based estimations of eastern Pacific Ocean yellow fin tuna TP (Olson et al. 2010).

Conclusion

The comprehensive SI analysis (599 measurements) provides a baseline for systematic investigations of PMNM trophic relationships and drivers of temporal and spatial variability in source isotopic signatures. Mixing model analysis provides support for mass-balance model estimates of the trophic importance of benthic primary production, demonstrating that BMA and BMI are vital components of a coral reef ecosystem with documented high macroalgal cover and high concentration of apex predator biomass (Friedlander and DeMartini 2002; Friedlander et al. 2008; Vroom and Braun 2010). The results suggest that the larger G. cuvier may be feeding at a higher TP above other apex predators and that TEF Δ15N may be relatively low in the PMNM. In addition, we report significant variability in primary producer SI values, the prevalence of a N2 fixation isotopic signature in all primary producers and qualitative evidence that undersampled non-N2 fixing primary producers are important contributors to the PMNM food web. The results improve our understanding of the trophic ecology of healthy coral reef ecosystems.