Seabird breeding success is closely related to the availability of adequate prey resources (Cury et al., 2011). As central place foragers, the geographic extent of a breeding seabird’s foraging activity is restricted by the need to return regularly to the nest site (Orians and Pearson, 1979). Thus, while seabirds may spend most of the year capable of traveling vast distances to feed where prey resources and foraging conditions are favorable (Gulka et al., 2023), during the breeding period they become dependent on the prey available within foraging range of the colony. This restriction makes breeding seabirds vulnerable to changes in the distribution, abundance, and composition of local prey resources (Boyd et al., 2017; Fayet et al., 2021).

In response to anthropogenic warming, many marine species have shifted their spatiotemporal patterns of occurrence, including numerous species that are important prey for breeding seabirds (Pinsky et al., 2013; Kleisner et al., 2017; Suca et al., 2021). These prey species, being largely heterothermic, are affected by ocean warming through direct physiological impacts; in contrast, seabirds react to ocean warming indirectly, reacting in response to the changing availability and distribution of their prey (Sydeman et al., 2015). Shifts in marine species distributions may be accelerated by marine heatwaves (MHWs) that temporarily – but often rapidly – bring anomalously warm conditions to a region, altering the distribution of thermal habitat (Lonhart et al., 2019; Jacox et al., 2020). While changes to a region’s marine community in response to warming are typically two-fold, involving both the “loss” of cold-water adapted species and the “gain” of warm-water adapted species (McLean et al., 2021), not all marine species are equally-suitable as prey for breeding seabirds (Harris et al., 2007; Smith and Craig, 2023). If changing ocean conditions lead to the rapid increase of less-suitable prey taxa, negative impacts on local seabird populations may follow (Descamps et al., 2022).

The suitability of marine taxa as prey for a particular seabird depends on varied traits (e.g., morphology, size, caloric value, lipid-content) that collectively summarize prey quality. When high-quality prey items predominate, seabirds can deliver more energy to chicks per foraging trip and breeding success tends to be high (Schrimpf et al., 2012). In contrast, the increased consumption of low-quality prey types may limit the reproductive success of marine predators (the junk food hypothesis; Romano et al., 2006; Österblom et al., 2008). Although seabirds may be able to compensate for variable prey conditions through behavioral adjustments like increasing foraging effort (Schrimpf et al., 2012), significant declines in the quality of available prey may exceed seabirds’ abilities to compensate (Ludynia et al., 2010; Wanless et al., 2005, 2023; Scopel et al., 2019).

The changing composition of Atlantic puffin (Fratercula arctica, hereafter puffins) chick diet has been proposed as a cause of recent declines in puffin reproductive success within the Gulf of Maine (GoM; Kress et al., 2016). A recent (2008–2010) thermal regime shift in the GoM was characterized by above-average ocean temperatures, followed by decreases in zooplankton biomass and cascading impacts across trophic levels (Johnson et al., 2018; Friedland et al., 2020b; Seidov et al., 2021). The timing of this regime shift closely aligns with both the observed decline in puffin reproductive success and changes in the composition of puffin chick diet (Kress et al., 2016; Scopel et al., 2019; Major et al., 2021). Concerningly, ocean temperatures within the GoM have continued to rise and the region is increasingly affected by MHWs (Pershing et al., 2018; Seidov et al., 2021). The result is that the GoM is now one of the most rapidly-warming parts of the global ocean, with significant declines predicted for many of the region’s species (Pershing et al., 2015, 2021).

When diet monitoring of GoM puffins began in the 1990s, Atlantic herring (Clupea harengus, hereafter herring) was among the most commonly observed prey species delivered (Diamond and Devlin, 2003; Diamond, 2021). Herring is an energy-dense species with a high lipid content, making it an important food item for many marine predators within the GoM (Lawson et al., 1998; Overholtz and Link, 2007; Spitz et al., 2010). However, both regional herring recruitment and occurrence in puffin diet have declined markedly in recent decades (Diamond and Devlin, 2003; Kress et al., 2016; Northeast Fisheries Science Center, 2018). In response, puffins have attempted to replace herring by provisioning chicks with alternative prey taxa like haddock (Meleanogrammus aeglefinnus) and rough scad (Trachurus lathami), neither of which were observed in GoM puffin diet before 2009. Despite the recent inclusion of these species into puffin diet, declining reproductive metrics suggest that these alternative prey species may not be adequate replacements (Scopel et al., 2019).

Puffin diet in the GoM has largely been assessed through the use of bill-load photography to identify the prey items provisioned by adults to chicks at the breeding colony (Kress et al., 2016; Scopel et al., 2019). This method is noninvasive and highly-effective at identifying major prey groups, although visual identification of prey is often limited to low-resolution taxonomic assignments (Gaglio et al., 2017). Furthermore, while bill-load photography may be suitable for estimating chick diet, it is unable to determine the diet of adults since adult puffins consume their prey at sea (Harris and Wanless, 2012). Evidence that seabird chick diet can be used as an effective proxy for adult diet is mixed (Wilson et al., 2004; Bowser et al., 2013) and the diets of adult seabirds, generally, are poorly understood compared to those of chicks (Barrett et al., 2007). However, since adult body condition is known to directly impact reproductive success, adult diet during the breeding period is likely equally as important to understand as chick diet (Erikstad et al., 1997; Barrett et al., 2012). Given the recent decline in GoM puffin reproductive success and continued climate-mediated shifts in the local forage fish community, reassessments of puffin diet and the methods used to determine this are warranted.

In this study, we aim to determine the diet of breeding puffins through the use of two complementary methods: traditional bill-load photography and molecular diet assessment using fecal DNA metabarcoding. From these two methods, we describe the prey types occurring in the diet of puffin chicks and breeding adults across incubation and chick-rearing during two anomalously warm breeding periods. We then examine differences in the composition of diets as determined through the different methodologies as well as between different age classes, breeding stages, and years.

During the 2021 and 2022 field seasons, researchers performed 250 hours of observation-based puffin diet surveys. A total of 692 photographed provisioning loads ( Table 1 ) were included in these analyses, representing more than 2,500 identifiable prey items. For molecular dietary assessments, researchers collected a total of 414 fecal samples from Matinicus Rock puffins. Of these, 176 (43%) amplified sufficient dietary DNA using 18S primers to be included in our assessment of broad metazoan taxa represented in puffin diet. Using 12S MiFish primers, 285 (69%) amplified to a sufficient depth to be included in our analyses of fish taxa consumed by puffins.

Using a combination of observational and molecular methods, we demonstrate significant variation in Atlantic puffin diet across ages, breeding stages, and years. This study provides the first formal assessment of puffin diet during the incubation period and is the first multi-year comparison of breeding puffin adult and chick diets using fecal DNA metabarcoding. Emerging, molecular methods produced similar estimates of chick prey consumption as traditional, observational methods for dominant prey types. Yet, metabarcoding identified twice as many fish prey species overall. Our study reveals that puffin diet in the GoM is both diverse and highly variable.

The use of molecular methods enabled us to examine puffin diet during the incubation period for the first time. We hypothesized that, lacking the demands of frequent chick-provisioning, adult puffins would be able to perform longer, more distant foraging trips during incubation. This could enable them to access prey resources not exploited during chick-rearing, similar to the strategy used by some thick-billed murres (Uria lomvia; Ito et al., 2010). While we did observe significant differences between diet composition between the incubation and chick-rearing stages, an examination of the taxa driving the observed dissimilarities suggests that these differences may have more to do with changes in local prey availability than different habitat selection. In 2022, for example, the RRA of both white hake and Atlantic herring was higher during the incubation stage than during chick-rearing. Both of these species are comparatively cold water adapted (Rose, 2005; Kleisner et al., 2017) and perhaps followed suitable thermal habitat away from the breeding colony as local waters exceeded key temperature thresholds. Other differences between periods may be explained by interspecific variation in the timing of fish transitioning from larval stages to the age-0 juveniles generally consumed by puffins. It must be noted that our sample sizes during incubation were small and were likely insufficient to detect all prey taxa consumed ( Supplementary Figure 1 ), so puffin diet during incubation is likely more diverse than we report here. Further work on seabird diets immediately before and after egg laying is encouraged since the prey availability and consumption during this time may have important impacts on reproductive success (Barrett et al., 2012).

Better known is how seabird diet during chick-rearing can influence reproductive output. However, comparatively few published studies have compared seabird adult and chick diets simultaneously (Wilson et al., 2004; Bowser et al., 2013; Fayet et al., 2021). Our results agree with the available literature in determining that the dominant prey types consumed by adults and chicks generally overlapped, but that adults consumed a greater overall variety of species. On an individual sample level, samples from adults had a greater mean species richness than those from chicks, while adults’ Shannon Index values were similar to those from chick samples. This may mean that adults “snacked” on more diverse prey types than chicks but generally relied on a similar number of species. While we found adult puffin diet to be highly diverse (28 fish taxa detected), more than 90% of their diet was comprised of the eight most common fish taxa each year.

An important finding of our study is that adult and chick diets were statistically distinguishable. Despite feeding on the same dominant prey types, the relative importance of these fishes in puffin diet varied by age class. Interestingly, the only previous comparison of puffin adult and chick diets in the GoM found their compositions to be generally similar (Bowser et al., 2013). However, the degree of dietary differentiation between age classes likely varies with prey conditions (Baird, 1991). Bowser et al.’s samples were collected during a year (2009) of anomalously high herring abundance where puffin adult and chick diets differed little because herring was detected in each sample. Likewise, in a multi-colony study of puffin foraging ecology (Fayet et al., 2021), locations where prey conditions and reproductive success were high (e.g., Wales) had highly similar adult and chick diets. During years of poor prey availability, as in our study, adults must be additionally selective about the prey items they provision to chicks (Burke and Montevecchi, 2009). Greater selectivity at these times likely leads to greater differences in chick and adult diets. This suggests that the use of chick diet as a proxy for adult diet could be acceptable in years of very high prey availability, but that diets may diverge greatly when prey conditions are less favorable.

As predicted by optimal foraging theory (Orians and Pearson, 1979), we found that adults appeared to prioritize higher-quality prey for chick-provisioning and fed on lower quality items themselves. Those species consumed more often by chicks, like sandlance, Atlantic saury, and mackerel (Scomber scombrus), were generally lipid-rich (5–10% lipids by wet mass; Spitz et al., 2010), large (≥ 2 bill lengths), or both. In contrast, adults consumed more small and low-lipid prey types like juvenile hakes and invertebrates (≤ 3.7% lipids by wet mass; Spitz et al., 2010). Interestingly, haddock were a greater component of chick diet than adult diet in 2021, despite being lipid-poor like most gadids (Harris and Hislop, 1978; Spitz et al., 2010). This may be because provisioned haddock tended to be large (x̄ = 2.1 bill lengths) and their mass could thus compensate for a low energy density. This highlights that prey quality is challenging to define using any singular metric and demonstrates the value of multiple, simultaneous methods of diet assessment in interpreting dietary data.

Relative to previous years of observational chick diet monitoring in the GoM (Kress et al., 2016; Scopel et al., 2019), the composition of puffin diet during the two years of this study was unusual. Haddock, consistently present in GoM puffin diet since it was first noted in 2009 (Bowser et al., 2013) was exceptionally scarce during 2022. Additionally, two typically uncommon species – rough scad and Atlantic saury – were provisioned in frequencies never before recorded at this colony (National Audubon Society’s Seabird Institute, unpublished data). Both rough scad and Atlantic saury favor warm waters (Collette and Klein-MacPhee, 2002) and may have followed rising temperatures into the GoM during the two anomalously warm years of our study (Mills et al., 2023). As the occurrence of herring in puffin diet has declined in association with warming ocean temperatures (Kress et al., 2016; Scopel et al., 2019), puffins in the GoM have demonstrated remarkable dietary flexibility; here, puffins appeared to exploit the infrequent, but opportune abundance of these two species. As generalist predators, puffins are capable of feeding on alternative prey types and diversifying their foraging strategies in order to mediate the effects of variable prey availability (Baillie et al., 2004; Schoen et al., 2018). The below-average puffin reproductive success during this study (2004–2020 mean: 0.65 chicks nest^-1; 2021–22 mean: 0.45 chicks nest^-1; National Audubon Society’s Seabird Institute, unpublished data), however, suggests that even the exploitation of these alternative prey types was insufficient to fully compensate for the poor prey conditions during the 2021–2022 MHWs.

In addition to the unprecedented abundance of Atlantic saury and rough scad in puffin diet, this study documents the occurrence of four species in puffin diet for the first time: Atlantic salmon, spotted hake, white mullet, and northern sennet. The appearance of salmon may be related to higher migration success of smolts following dam removals in Maine and regulatory changes at remaining dams (Stevens et al., 2019). The occurrence of salmon in the diet of two 2021 puffins during June is likewise well-timed with the spring seaward movements of smolts (Kocik et al., 2009). The presence of the other species is likely related to warming ocean conditions and shifting thermal habitat. Spotted hake, white mullet, and northern sennet are typically uncommon north of Cape Cod, being more abundant within the warmer Mid-Atlantic Bight (Collette and Klein-MacPhee, 2002). Warming conditions – exacerbated by MHWs – have led to range shifts for many marine species and the general tropicalization of the U.S. Northeast Shelf (Kleisner et al., 2017; Friedland et al., 2020a). Our detection of these novel species in puffin diet demonstrates the potential efficacy of using metabarcoding techniques and generalist, marine predators to monitor changing marine communities in response to global climate change (Sydeman et al., 2017; Lenoir et al., 2020).

The greater taxonomic resolution offered by metabarcoding also enabled us to identify more fish taxa in puffin fecal samples than with visual observations alone. Precise visual identification of juvenile hakes from bill load photography is not possible, and observations must be grouped as “hake spp.” in these assessments. In contrast, we were able to identify five unique species of “hakes” (fourbeard rockling and white, red (Urophycis chuss), silver (Merluccius bilinearis), and spotted hakes) using fecal DNA. With molecular identification, we observed changes in the composition of “hake spp.” over our study; whereas white hake predominated during incubation and throughout 2022, hakes during chick-rearing in 2021 were overwhelmingly fourbeard rockling. While seemingly trivial, closely related species may have different thermal tolerances and responses to changing ocean conditions (Rose, 2005); although MHW conditions dominated both years of our study, ocean conditions were notably warmer in 2021 (Mills et al., 2023) and may have contributed to the scarcity of the less heat-tolerant white hake in puffin diet that year. Beyond thermal tolerances, morphologically-similar species can also have different lipid contents, resulting in variable energetic gain for their predators (Spitz et al., 2010). The use of diet assessment methods capable of such high-level identification can therefore be valuable for understanding the complex relationships between marine predator diet, reproduction, and ocean warming.

Despite these advantages, metabarcoding techniques are not a panacea and cannot replace certain types of data collected using traditional methods. Most obviously, metabarcoding is not capable of estimating the total mass of prey consumed or the size or age class distributions of prey items. Whereas stomach flushing revealed that common murre (Uria aalge) adults tend to consume smaller, younger age classes of sandeels (Ammodytes spp.) than they provision their chicks (Wilson et al., 2004), the isolated use of molecular methods would have been unable to discern this. Furthermore, some interpretation of molecular diet assessments are necessary to account for potential secondary and incidental ingestion (Sheppard et al., 2005). We detected the occurrence of 17 unique metazoan groups in puffin diet but this total includes many taxa unlikely to have been consumed directly by puffins due to habitat (e.g., terrestrial winged insects) or size (e.g., copepods). We thus relied heavily on concurrent visual observations of prey deliveries to determine which invertebrate taxa were most likely consumed by puffins directly and which were likely consumed by puffin prey. Since we detected various differences in adult and chick diets, we note that this method is likely imperfect and that even our dual-methodology may be inadequate to fully assess adult puffin invertebrate consumption. Although our decisions regarding which taxa represent intentional consumption were largely substantiated by the available literature on both puffin (particularly Piatt, 1987; Falk et al., 1992; Bowser et al., 2013; Harris et al., 2015) and forage fish diet (Bowser et al., 2013; Suca et al., 2018), it is possible that adult puffins exploit invertebrate prey resources more broadly than currently thought. Simultaneous diet sampling across multiple trophic levels (i.e., diet studies of seabirds and their prey; see Bowser et al., 2013) will doubtless aid in the interpretation of metabarcoding-derived diet data.

Generally, though, we believe our dual-method strategy for assessing puffin diet confirms the utility of fecal DNA in supplementing traditional methods of diet assessment. A key finding of our research is that the estimated relative contribution of a prey taxa to puffin diet was found to be strongly correlated when estimated through traditional and molecular methods, as confirmed recently elsewhere (Clucas et al., 2024). As in other metabarcoding studies, the correlation was strong though imperfect due to various factors such as the differential survival of DNA during digestion, interspecific variation in the amount of DNA per gram of tissue, primer binding biases, and PCR stochasticity (Deagle and Tollit, 2007; Thomas et al., 2016; Alberdi et al., 2019). Despite this, the estimated relative importance of most dominant prey types was highly similar between methods and we believe this highlights the efficacy of using metabarcoding to quantitively assess seabird diet. We believe the relative composition of puffin diet may be reliably estimated using the RRA from fecal DNA metabarcoding methods, including for ages and breeding stages not previously possible to assess noninvasively.

Further monitoring of puffin diet within the GoM will provide valuable information on how puffins are adapting to changing prey-resources. The loss of puffins’ dominant prey species in other regions has very rapidly led to widespread breeding failure (Barrett et al., 1987; Miles et al., 2015) but declines in reproductive success within the GoM have so far been milder (Scopel et al., 2019). The high diversity of prey taxa available in the region has likely served as a buffer against the decline of herring and other cold water-adapted species (the portfolio concept; Schindler et al., 2015). This diversity and the opportunistic nature of puffins in exploiting irregularly-occurring prey resources may help this population be resilient to ongoing and projected oceanographic changes within the GoM.

The data presented in the study are deposited in the figshare online repository, https://doi.org/10.6084/m9.figshare.25365754.v1.

The animal study was approved by U.S. Fish and Wildlife Service personnel at Maine Coastal Islands National Wildlife Refuge. The study was conducted in accordance with the local legislation and institutional requirements.

WK: Software, Writing – review & editing, Writing – original draft, Visualization, Methodology, Formal analysis, Conceptualization. GC: Software, Writing – review & editing, Resources, Methodology, Formal analysis, Data curation, Conceptualization. DL: Writing – review & editing, Supervision, Funding acquisition, Conceptualization.