Effects of Atlantic butterfish ( Peprilus triacanthus ) in diets of Common Terns ( Sterna hirundo ): a case study of climate change effects in the Gulf of Maine

. Climate change and associated shifts in marine prey communities can alter food availability for foraging seabirds. This issue is illustrated in the Gulf of Maine by the northward shift of Atlantic butterfish ( Peprilus triacanthus ; hereafter butterfish) and their subsequent increase in seabird diets. Here, we examine effects of butterfish in diets of the Common Tern ( Sterna hirundo ), a threatened species in this region. Our objectives were to evaluate butterfish suitability for tern chick diet through observing handling time and feeding success, to examine effects of butterfish on tern chick growth, and to explore signs of adaptive foraging in adults. The diet and growth of Common Tern chicks were studied for three breeding seasons on Seavey Island, New Hampshire, USA in the Gulf of Maine. Prey items were identified during feedings and were grouped into five prey categories: butterfish, herring, hake, other fish, and invertebrates. Chicks were weighed to calculate growth rate, which was examined as a response to diet. Across prey categories, butterfish were handled by chicks for the greatest amount of time (P < 0.001) but were swallowed the least (P < 0.001). Furthermore, chick growth rates were negatively associated with proportion of butterfish in the diet (P < 0.001). There was significant variation in chick diet across different nests in two of the three years studied (P < 0.05 in 2018 and 2019), and some breeding pairs were never observed provisioning butterfish to their young. Although the mechanism underlying individual specialization is not known, chicks that are fed fewer butterfish are at an advantage. The provisioning (or not) of unsuitable prey is particularly important because seawater warming in the Gulf of Maine is expected to increase, and butterfish are likely to become even more prevalent in seabird diets.


INTRODUCTION
Climate change affects physical conditions of marine systems, eliciting community responses in which some species may shift their distribution or phenology (Doney et al. 2012, Poloczanska et al. 2013, Sydeman et al. 2015. However, not all species in a community respond in the same manner or magnitude, and asymmetries in species' responses to climate change can alter predator-prey interactions through changes in size, location, and overlap of an organism's spatial range (Selden et al. 2018). For species that are central place foragers, these shifts in prey distribution can increase the distance between the predator's foraging zone and its desired prey, ultimately altering food availability (Durant et al. 2007). Some species adapt to these changes by replacing historically preferred prey with different species or by adjusting their spatial range according to changes in their environment (Grémillet and Boulinier 2009).
Instances of predator-prey mismatch as a response to shifting prey communities have been observed in the Gulf of Maine (GOM; Staudinger et al. 2019), which is warming faster than 99% of the world's oceans (Pershing et al. 2015). In these prey communities, climate-driven ecosystem changes have resulted in more temperate species and fewer subarctic organisms (Pershing et al. 2021). One such change to the GOM prey community can be seen in distributional shifts of Atlantic butterfish (Peprilus triacanthus; hereafter butterfish), a subtropical to temperate fish that is found in Northwest Atlantic waters ranging from Florida, USA to Nova Scotia, Canada (Adams et al. 2015). Butterfish is vulnerable to distributional shift because it is a highly mobile habitat generalist with broad larval dispersal (Hare et al. 2016) and migrates in response to fluctuating water temperature (Hudson and Peros 2013). Consequentially, this warm-water fish has shown significant northward shift in mean latitude in recent decades (Suca et al. 2021), including range expansion into the GOM (Adams 2022). Butterfish were detected in the GOM as early as 1982 (Adams 2022). However, more recently, they are increasing in diets of seabirds within this region (Kress et al. 2017) and are replacing historically preferred prey of seabirds (Scopel et al. 2019). Despite being captured more frequently, butterfish may prove to be unsuitable prey for seabirds because of its dorsoventrally deep-bodied shape, which makes it difficult for gape-limited seabirds to swallow (Kress et al. 2017). This issue is of particular concern for foraging seabirds in the GOM because, in coming decades, habitat suitability for butterfish is predicted to be high in this region (McHenry et al. 2019).
Here, we explore effects of butterfish in the diet of Common Terns (Sterna hirundo) at a breeding site in the GOM. Common Terns are the most widely distributed tern in the GOM and in North America. These birds were overharvested and displaced from their colonies during the 20th century; today, they are considered a Regional Species of Greatest Conservation Need in the U.S. Northeast region (Terwilliger Consulting 2015). Breeding Common Terns are central place foragers constrained by foraging distance away from their young (with a typical clutch size of 2-3 eggs; Arnold et al. 2020). In the GOM, they capture a variety of prey from foraging areas surrounding their breeding colonies, including small and juvenile fishes (particularly species of herring and hake [Clupea, Alosa, Urophycis, and Merluccius species]) and invertebrates such as insects, euphausiids, and squid (Hall et al. 2000, Arnold et al. 2020. Understanding effects of climatemediated changes on Common Tern prey is particularly important for the protection of this management-dependent seabird, as prey availability is an essential component underlying seabird reproductive success during the breeding season (Schaffner 1986, Cairns 1988, Suddaby and Ratcliffe 1997. Effects of climate change and overfishing on prey availability, as well as increased butterfish in the diet, are specifically cited as threats to breeding terns in New Hampshire, USA (Carloni 2015).
We examined suitability and effects of butterfish in diets of Common Terns to evaluate current and potential future impacts of increasing butterfish abundance with warming temperatures in the GOM. Our specific objectives were: 1. To evaluate the suitability of butterfish for chick consumption, using handling time and feeding success as measures of prey suitability. We hypothesized that due to their deep-bodied morphology, butterfish would be difficult to handle and ultimately less likely to be successfully swallowed compared to other prey items, rendering them unsuitable prey for tern chicks.

2.
To examine effects of butterfish on tern chick growth. We expected that increased rates of butterfish provisioned to tern chicks by their parents would negatively affect chick growth.

3.
To explore variability in butterfish provisioning among breeding pairs of adults. We examined diet composition across nests and hypothesized that variability in the proportion of butterfish provisioned may be an indicator of an adult's capacity to respond adaptively to the increasing presence of butterfish in the GOM prey community.

Study area
We collected data from Common Terns at a mixed-species tern nesting colony on Seavey

Common Tern chick diet
Feeding watches were made either in-person with binoculars from observation blinds or remotely using video footage recorded from an AXIS P5635-E Mk II PTZ Network Camera. Observations occurred between 5:30 and 21:30. Chicks monitored during feeding watches were banded and regularly marked on the breast with colored ink, allowing us to identify individuals and their corresponding nest identity (ID). We could not consistently identify individual adults during feeding watches, so our diet data were collected at the nest level, representing each pair of adults. The majority of nests were monitored for ≥ 5 h over the course of the breeding season. For nest-level analyses, we removed seven nests (of 53 total) that were observed for < 5 h, finding that the low watch times resulted in feeding rate outliers.
For each chick feeding event, we visually identified prey to the lowest taxonomic resolution possible. Prey items recorded were grouped into five categories: (1) butterfish, (2) (5) invertebrates (primarily euphausiids, insects, and squid). Thirty-five percent of prey items were unidentifiable and were excluded from analyses. Prey size was quantified by visually estimating the length of prey relative to that of the adult tern's bill, using units of "bill length." To assess the suitability of butterfish as prey for Common Tern chicks, we recorded handling time and feeding success. Handling time was defined as number of seconds that a chick held a prey item in its bill before it was swallowed, discarded, or stolen.
Handling times were included in analyses regardless of whether prey were swallowed. Feeding success was defined as the proportion of each prey category that was successfully swallowed by chicks; discarded or stolen prey were considered unsuccessful feedings. To test for differences in both handling time and feeding success across prey categories, we used analysis of variance with post-hoc Tukey honestly significant difference (HSD) tests. We used linear regression to assess the relationship between prey size and handling time within each prey category. Statistical analyses were conducted in R (version 4.0.2; R Core Team 2021) and JMP (JMP Pro 15.0.0).

Common Tern chick growth rates
We weighed individual chicks at least twice between 3 and 15 days of age, the period over which chick weight increases linearly and a linear growth rate could be calculated (Arnold et al. 2016). For nests with more than one chick, we used the growth rate of the fastest growing sibling in analyses, as one chick typically received most feedings, and that chick's growth was therefore more representative of the composition of the diet. Furthermore, in 60% of nests with more than one chick, one sibling tended to die quickly (within one week of hatching in our study population), limiting the potential to collect diet and growth data for multiple chicks in one nest. For growth-diet comparisons, we used feeding data when chicks were between 0 and 15 days old, as these feedings were most relevant to the period of linear chick growth. To assess effects of butterfish in diets on chick growth while accounting for overall feeding rates, we conducted a least-squares regression of chick growth rate, including the effects of proportion of butterfish in the diet, overall feeding rate, and the interaction between those variables. Although we also measured juvenile survival (to 15 days of age) at our study site, this metric was highly influenced by nondietary factors, including severe weather and predation events. We therefore chose to focus our investigations on chick growth rate as the response metric most immediately influenced by shifts in diet.

Variability in butterfish provisioning
We examined variability in number of butterfish in diets across years and across nests within each year of study. For these analyses, we grouped prey items into two categories: butterfish and all other prey. We compared the number of butterfish relative to all other prey provisioned across the three years of study using a χ 2 test. To examine variability in butterfish provisioning by different nesting pairs of adults, we compared number of butterfish and other prey by nest ID within each of the three years using a two-tailed Monte Carlo estimate of Fisher's exact test (Mehta and Patel 2013).

RESULTS
We conducted

Variability in butterfish provisioning
The number of butterfish relative to all other prey in chick diets varied across the three years (χ 2 = 54.7, degrees of freedom = 2, P < 0.01). We found a significant relationship between diet and nest ID in 2018 (P < 0.01) and 2019 (P = 0.04), but not in 2017 (P = 0.57) using a Monte Carlo version of Fisher's exact test.

DISCUSSION
Common Tern chicks spent substantial time handling butterfish but infrequently gained nutritive rewards, with only 21% of butterfish being successfully swallowed (Fig. 1). Butterfish was the only prey category for which prey size had no significant effect on handling time.

Fig. 1. Suitability of Atlantic butterfish (Peprilus triacanthus)
and other prey categories in the diet of Common Tern chicks (Sterna hirundo) in the Gulf of Maine as measured by handling time (top; logarithmic scale) and feeding success (bottom). Different letters signify significant differences among prey categories according to Tukey HSD tests (P < 0.001 for both handling time and feeding success comparisons).
This result indicates that difficulty handling and ingesting butterfish is not a function of total length, but is likely driven by dorsoventral body depth, which, although not directly measured here, typically exceeded the gape of tern chicks across the range of butterfish sizes provisioned by adults. We also observed occasions during which chicks ignored other prey items offered to them while they were preoccupied handling butterfish, compounding nutritive costs of butterfish provisioning. Unconsumed butterfish were often discarded near the nest, where they were either picked up by a neighbor or left long enough to desiccate and decay.
Chicks that were fed higher proportions of butterfish had lower growth rates, likely due to the low probability of successful feeding and the associated energetic costs of long handling times.
Although we expected to find that increased overall feeding rates would correlate with higher growth rates in chicks, we found this relationship to be positive but not significant. However, when considering the significant interaction between proportion of butterfish and overall feeding rate (Fig. 2), we found that increased feeding rates were associated with increased growth rate as expected, but only when proportions of butterfish in diets were low. When proportions of butterfish in diets were high, there was no association between feeding rate and chick growth. The association of butterfish provisioning with reduced growth rates in Common Tern chicks is particularly concerning, given the strong linkage between chick growth and survival in seabirds (Maness andAnderson 2013, Kress et al. 2017). Furthermore, post-fledging body mass has been linked to subadult survival in this species, and if chicks fail to put on weight prior to migration, we may see fewer young, prospecting birds returning to this colony in the future (Braasch et al. 2009).

Fig. 2.
Contour plot of chick growth rate (g/day) as it relates to overall feeding rate (feedings/h) and proportion of Atlantic butterfish (Peprilus triacanthus) in the diet of Common Tern chicks (Sterna hirundo) in the Gulf of Maine.
Given that butterfish appear to be highly unsuitable prey for tern chicks, the question remains as to why tern adults feed their chicks this fish. We speculate that when encountering juvenile butterfish on foraging grounds, adult terns may capture this fish because it appears similar to other common, high-quality prey such as juvenile herring and hake, which are similar in length and also laterally compressed. However, once captured, we would presume that the deep-bodied morphology of butterfish would be apparent to an adult tern. This mismatch led us to question whether adults are sensitive to prey suitability.
We quantified variability in butterfish provisioning within the sampled population and found that number of butterfish in diets varied significantly among nesting pairs in 2018 and 2019, although not in 2017 (potentially due to smaller sample size in that year; Fig. 3). This finding highlights that adult terns breeding on the same island during the same season feed their chicks significantly different diets. In all three years, there were breeding pairs that were never observed provisioning butterfish to their chicks, even in 2018, when the relative abundance of butterfish in the diet of terns at this colony was highest. Adult terns adjust the size of prey provisioned when foraging, depending on nutritive needs of nestlings (Quiring et al. 2021), and our results suggest they are species-selective as well. However, these signs of individual specialization may not signal active butterfish avoidance, but could be driven by numerous factors, including other characteristics of the prey field (e.g., the distribution and abundance of high-quality prey) as well as characteristics of individuals birds (e.g., age, sex, physiological fitness, and site fidelity; Phillips et al. 2017, Nisbet et al. 2020. While the actual mechanism of dietary variation at this colony is a topic of ongoing investigation, chicks of terns that capture fewer butterfish, by whatever mechanism, are at an advantage. This consequence provides some means for adaptive response in the face of shifting prey communities in the GOM, contingent on continued availability of suitable prey into the future.

CONCLUSION
The effect of butterfish on Common Tern chicks described here highlights the significant role that prey species composition plays in the reproductive outcomes of terns. Continued monitoring of tern diets will be key to evaluating how and whether these seabirds are able to respond to changes in their prey communities, which are simultaneously affected by climate change and commercial fishing in the GOM (Pershing et al. 2021). Seabird diet data such as those presented here can fill gaps in fish stock assessments (Sydeman et al. 2022), including GOM fisheries for butterfish and other common prey species (e.g., Atlantic herring; Scopel et al. 2018, Deroba et al. 2019. We additionally suggest that seabird diet data can contribute to a more holistic approach to fisheries management (e.g., ecosystem-based fisheries management; Pikitch et al. 2004) that incorporates the needs of predators such as seabirds (Townsend et al. 2019). Through such an approach, juvenile fishes that are important both commercially and ecologically (Pikitch et al. 2014) can be monitored and protected to sustain marine ecosystems as a whole (Pikitch et al. 2004).