Tern diet in the UK and Ireland: a review of key prey species and potential impacts of climate change

Transcription

1 Tern diet in the UK and Ireland: a review of key prey species and potential impacts of climate change October 2017 Elizabeth Green Species and Habitats Officer, RSPB The project "Improving the conservation prospects of the priority species roseate tern throughout its range in the UK and Ireland" is supported by the LIFE Programme of the European Union. LIFE14 NAT/UK/394 ROSEATE TERN

2 Contents Background... 1 Introduction... 1 The diet of breeding terns in the UK, Ireland and southern North Sea... 2 UK, the North Sea... 3 West Scotland... 7 The Irish Sea... 7 The Celtic Sea... 9 The Southern North Sea... 9 The Wadden Sea, south-eastern North Sea Summary The ecology and status of sandeels, sprat and herring Sandeels Sprat Herring Mapping prey hotspots and tern resilience Prey hotspots Resilience and vulnerability of tern colonies to prey shortage Impacts of climate-driven environmental change on sandeels, sprat and herring Observed environmental change Predicted future environmental change Summary and recommendations for fisheries management Acknowledgements References... 36

3 Background In order to improve the status of priority species, pressures on populations must be identified and effectively managed. For the roseate tern, a species of high conservation concern in the UK, this requires an improved understanding of the key issues affecting populations in both the breeding and wintering grounds. One such issue is the availability of high-energy food during the breeding season, when terns must collect enough food to sustain themselves and raise their chicks. Due to the small roseate tern population in the UK (~100 pairs) and Ireland (~1,400 pairs), little research has been carried out on the diet of this species. However, the diets of other tern species in this region and the southern North Sea have been more extensively studied and, although inter-specific variation in diet exists, key prey species are fairly consistent across tern species. Improving our understanding of the diets of Arctic, common, little and Sandwich terns will also support the conservation of these species, for which the UK holds internationally important breeding populations. This report, carried out as part of the EU-funded Roseate Tern LIFE Project, reviews current knowledge regarding the diets of the five tern species that breed in the UK and Ireland and the ecology of their key prey species. The impacts that ongoing climatedriven environmental change may have on key fish species are also discussed, and recommendations are made that aim to safeguard these prey populations and the seabirds they support. Introduction Food availability is an important driver of seabird breeding success (Cairns, 1988). High food availability can increase productivity by improving survival or growth rates of chicks or enabling adults to raise a greater number of chicks (Martin, 1987; Monaghan et al., 1989; Robinson and Hamer, 2000). The energy content of food is also a critical factor, and varies between different prey species (Harris and Hislop, 1978) but also within species depending on body condition. Indeed, years of poor body condition of prey have been associated with poor seabird breeding success (Wanless et al., 2005; Österbrom et al., 2006; Österbrom et al., 2008). Such bottom-up effects of forage fish on predators are likely to be strongest when the predator is a specialist that feeds on one or a small number of fish species (Cairns, 1987; Engelhard et al., 2014). One such group of piscivores is the terns. Five tern species breed in the UK: the Arctic tern Sterna paradisaea, common tern Sterna hirundo, little tern Sternula albifrons, roseate tern Sterna dougallii and the Sandwich tern Sterna sandvicensis. Terns target a relatively small number of prey species and generally behave as single prey loaders (although see Dunn, 1972), making them particularly sensitive to changes in food availability and quality (Furness and Tasker, 2000). Tern chicks are generally fed on high energy prey, with low quality species or small individuals being consumed by adults during foraging trips (Dunn, 1972; Ewins, 1985; Wilson et al., 2004; Perrow et al., 2010; Perrow et al., 2011b). However, during periods of low availability of their preferred prey, chicks may be fed on less calorific items. High energy prey tend to be marine fish which have a high protein and fat content, such as sandeels and clupeids (Harris and Hislop, 1978; Massias and Becker, 1990), whereas low energy prey contain more indigestible material and include crustaceans, such as shrimp, and freshwater fish, such as the three-spined stickleback (Massias and Becker, 1990). Experimental studies have shown that chicks fed on a high quality prey (herring) grow quickly while chicks fed on low quality prey (shrimp or sticklebacks) suffer poor growth rates and in some cases even lose weight (Massias and Becker, 1990). This highlights how crucial it is for terns to have access to an abundant supply of high quality prey during the breeding season. 1

4 Tern diet is likely to be affected by foraging range, as different prey species associate with different areas and habitats (Reay, 1970; Whitehead, 1985; Araujo et al. 2000; Wright et al., 2000). The tern with the shortest foraging range is the smallest-bodied of the five species, the little tern, while the largestbodied species, the Sandwich tern, has the largest foraging range (Eglington and Perrow, 2014). Thaxter et al. (2012) reviewed studies of foraging distances of seabirds and reported that the maximum foraging range of little terns was 11 km, while that of Arctic, common and roseate was 30 km and Sandwich terns were reported to travel up to 54 km. The mean foraging ranges estimated in the same study were much lower, at 2.1 km for little terns, 7.1 km for Arctic terns, 4.5 km for common terns, 12.2 km for roseate terns and 11.5 km for Sandwich terns. Moderate confidence was associated with the estimates of Arctic, common and Sandwich terns, whereas estimates of little terns and roseate terns were described as low confidence due to a lack of direct measurements (Thaxter et al., 2012). There is substantial variation in estimates of foraging distances between years, colonies, individuals, tidal conditions (Steinen et al., 2000; Perrow et al., 2004; Paiva et al., 2008; Perrow et al., 2010) and even reviews; see Eglington and Perrow (2014) for a comprehensive summary. Therefore it may be inappropriate to consider species as being either strictly inshore or offshore foragers, except perhaps for the little tern (inshore). Diet composition is also influenced by the time of day, the tide and the weather (Veen, 1977; Frick and Becker, 1995; Wendeln, 1997; Stienen et al., 2000; Morris 2012). For example, some diet studies have found that clupeids are most frequently caught in the mornings and evenings, while sandeels are most frequent during the middle of the day (Stienen et al., 2000; Morris 2012). This corresponds with greater availability of clupeids in the early morning and evening and sandeels during daylight hours to surfacefeeding birds, due to differences in the timing of vertical movements in the water column (Blaxter and Parrish, 1965; Stienen et al., 2000; Falkenhaug and Dalpadado, 2014). During periods of poor weather, marine prey may become inaccessible to foraging terns, resulting in greater dependence on lower quality prey such as sticklebacks caught in more sheltered water bodies (Wendeln, 1997; Stienen et al., 2000). The relative importance of key prey species also varies between tern species and regions. This variation means some species and areas are particularly vulnerable to changes in prey populations, which can be caused by bottom-up processes (such as climate-driven changes in the zooplankton community), topdown processes (such as fishing mortality) and the interactions between these (Engelhard et al., 2014; Lynam et al., 2017). It is important to identify these vulnerabilities in order to gain a greater understanding of local pressures on tern colonies, thereby informing development of appropriate colony-specific management plans. The diet of breeding terns in the UK, Ireland and southern North Sea Below is a review of geographic and inter-specific variation in tern diet in the UK, Ireland and the southern North Sea. Although effort was made to collect information from all areas around the UK, there is a lack of information relating to the colonies along the southern coast of England. Figure 1 shows a map of seas and locations around the UK, Ireland and southern North Sea referred to in the following text regarding tern diet. 2

5 Figure 1. Map of locations where tern diet has been studied and is referred to in the text: SH = Shetland Islands; FF = Firth of Forth; FI = Farne Islands; CQ = Coquet Island; SA = Saltholme, Teesside; GP = Gibraltar Point; GW = Greater Wash; BP = Blakeney Point; WI = Winterton; ND = North Denes; HI = Horse Island, Firth of Clyde; SK = Skerries, Anglesey; CB = Cemlyn Bay, Anglesey; YF = Ynys Feurig; RO = Rockabill Island; LI = Lady s Island Lake; SC = Isles of Scilly; HA = The Haringvliet; GR = Griend; MO = Minsener Oldeoog; BS = Banter See. UK, the North Sea Shetland Islands Sandeels are particularly important for terns at the Shetland Islands due to low availability of other small, energy-rich fish species in the area (Furness and Tasker, 2000), and are the main prey fed to common and Arctic tern chicks (Furness, 1982; Ewins, 1985; Monaghan et al., 1989). Inter-annual variation in Arctic tern productivity at Shetland can be largely explained by variation in sandeel biomass (Suddaby and Ratcliffe, 1997), and a low availability of 0-group sandeels during chick-rearing can result in total breeding failure (Monaghan et al., 1992). Sandeels are also crucial for adult Arctic terns at Shetland: a decline in the sandeel portion of adult diet from 100% to 20% between the early and late 1980s coincided with the number of breeding pairs declining by 70% (Bailey et al., 1991). However, the extreme vulnerability of Arctic terns to fluctuations in sandeel populations is not universal to all tern species at Shetland. Common terns in this area appear to be more capable than Arctic terns of targeting alternative prey species, thereby buffering the effects of variation in sandeel availability. For example, during a period of low sandeel availability at Shetland in 1988, ~80% of prey items fed to common tern chicks were saithe, with large sandeels making up the remainder (Uttley et al., 1989). Although saithe have a fairly low calorific value (Harris and Hislop, 1978), this resulted in common tern chicks achieving growth rates similar to those observed under good feeding conditions (Uttley et al., 1989). Arctic terns, however, continued to feed their chicks predominantly on small sandeels (<8 cm) which have a very low energy content (Harris and Hislop, 1978), and as a result growth rates were very poor (Uttley et al., 1989). Although during the study neither species fledged any young, breeding 3

6 failure of common terns was caused by predation rather than a lack of food (Uttley et al., 1989). In contrast, Arctic tern chicks died from nest desertion (30%) or starvation/exposure (70%), suggesting the likely cause of breeding failure in this species was low food availability (Uttley et al., 1989). Additional evidence of the sensitivity of Arctic tern productivity at Shetland to sandeel availability was demonstrated by comparing chick diet and breeding success at Garthness, in South Shetland, with that at Coquet Island (Monaghan et al., 1989). During the early chick rearing phase, Arctic terns at both locations fed their chicks on a similar proportion of sandeels (66-68%), with the remainder of the diet being mostly saithe on the Shetland Islands and sprat on Coquet (Monaghan et al., 1989). However, at Shetland there was a shortage of intermediate-sized sandeels and chicks were fed predominantly on very small sandeels ( 4 cm). This, in combination with a lack of high-energy alternative prey, resulted in chicks at Shetland having a lower growth rate and survival rate than those at Coquet (Monaghan et al., 1989). Firth of Forth Although sprat are largely absent from the Shetland Islands, a geographically isolated population exists further south in the Firth of Forth and supports a relatively large number of breeding common terns (Jennings et al., 2012; ICES, 2013a). However, over the past 50 years the local sprat stock has fluctuated markedly in relation to fishery pressure. A local sprat fishery, with a relatively large harvest given its size, operated on the north shore of the Firth of Forth until The stock collapsed in the early 1980s (Fernandez et al., 2005; cited by Jennings et al., 2012), and the fishery was closed in The stock has since recovered but the fishery has remained closed (Fernandez et al., 2005; cited by Jennings et al., 2012; Jennings et al., 2012). Sprat are the main prey of common terns in the Firth of Forth (Jennings, 2012) and these stock fluctuations caused substantial variation in breeding success at common tern colonies in the area (Jennings et al., 2012). Initial operation of the sprat fishery coincided with an increase in breeding numbers of common terns, with the number of pairs almost doubling between 1969 and 1972 (Jennings et al., 2012). The decline in landings during the early 1980s coincided with a decline in the breeding number of common terns, which remained at a relatively low level until 1993, 8 years after the sprat fishery closed (Jennings et al., 2012). The breeding numbers subsequently recovered and since then have been higher than they were during the initial high harvest period of the sprat fishery (Jennings et al., 2012). This demonstrates that terns are sensitive to fishing pressure and recovery of colonies following stock collapse of their prey can take many years (Jennings et al., 2012). Farne Islands Sandeels are less crucial for tern colonies south of the Shetland Islands, due to the presence of highenergy alternative prey, but they remain a major prey group in many areas and are particularly important for Arctic terns. A three year study at the Farne Islands in the early 1960s found that, while Arctic terns and common terns fed on the same prey species, Arctic terns fed their chicks less frequently on clupeids (sprat and herring) and gadoids and more frequently on sandeels (Ammodytidae) and marine invertebrates than their congenerics (Pearson, 1968). Common terns fed their chicks only a slightly lower proportion of clupeids than sandeels (Pearson, 1968). Pearson (1968) also considered the relative contribution of each prey group to total prey weight, as the most frequent prey in the diet does not necessarily contribute the most weight to overall food consumed (Wetherbee et al., 2004). Although sandeels were the most frequent prey fed to Arctic and common tern chicks, they accounted for less than a quarter of total prey weight for both species (Pearson, 1968). Conversely, clupeids contributed >60% of prey weight for both tern species, despite being fed less frequently than sandeels. However, 4

7 sandeels were the most important prey for Sandwich tern chicks in terms of both frequency and total prey weight (Pearson, 1968). Coquet Island Terns at Coquet have access to both sandeels and clupeids (sprat and herring), although inter-annual variation in diet composition suggests the availability of these two prey types can vary substantially between years (Langham, 1968; Morris, 2012). The relative importance of sandeels and clupeids and the size of prey fed to chicks varies between tern species (see below), but the patterns of increasing or decreasing proportions of a given prey type appear to occur in unison across all tern species. This suggests tern diet is heavily influenced by prey availability. For example, in 1965 clupeids (mostly sprat) and a small number of gadoids (such as whiting) were the most frequent prey fed to chicks of all tern species present on the island (Arctic, common, Sandwich and roseate; Langham, 1968). In the following year the proportion of sandeels in chick diet increased for all tern species and became the predominant prey for Arctic, common and roseate terns, although Sandwich terns continued to feed their chicks on a greater number of clupeids (Langham, 1968). Research carried out on Arctic and common terns since the 1990s has found that sandeels have consistently been the most frequent prey fed to chicks of both species (Robinson and Hamer, 2000; Robinson et al., 2001; Morris, 2012; Robertson et al., 2014, 2016). However, high energy years (years of high energy delivery rate and/or high energy per feed) are associated with a greater proportion of clupeids and larger prey in the diet (Morris, 2012). This is probably due to clupeids having more energy per length than sandeels (Hislop et al., 1991). General differences in diet between tern species are apparent. As in other areas, Arctic terns at Coquet feed their chicks on a greater proportion of sandeels and fewer clupeids than common, roseate or Sandwich terns (Langham, 1968; Robinson et al., 2001; Morris, 2012; Robertson et al., 2014). Only in 2011 did Arctic terns feed their chicks fewer sandeels than other tern species (Robertson et al., 2014). However, when only using data from the linear growth phase of chicks, Robertson et al. (2016) found that Arctic terns fed their chicks a higher proportion of sandeels than common terns. Although terns at Coquet may have access to a greater diversity of prey than terns at Shetland, variation in prey availability is still an important driver of breeding success. For example, between 2006 and 2010, sandeels (identified as A. marinus and A. tobianus) comprised 71% of the common tern chick diet and 83% of Arctic tern diet, while clupeids contributed most of the remainder (Morris, 2012). During this period there was substantial inter-annual variation in breeding success, which has been linked to variation in sandeel availability (Morris 2012). Low availability of sandeels in 2008 corresponded with a poor breeding season, whereas both 0-group sandeel abundance and tern breeding success were higher in 2009 (Morris 2012). In 2010 breeding was good for common terns, but not for Arctic terns. This coincided with a higher availability of large 1-group sandeels, which were probably too large for Arctic tern chicks but were suitable for common tern chicks (Morris 2012). Thus, even in areas where multiple prey species are found, terns remain sensitive to changes in availability of their major prey. Saltholme, Teesside As observed at other sites along the north-eastern and eastern coasts of England, terns at Saltholme have access to a range of prey types. In 2009 common terns fed their chicks predominantly on clupeids (mostly sprat), with the remainder of the diet consisting of sandeels, gadoids and sticklebacks (Perrow et al., 2010). Compared to observations of common terns at Blakeney Point (Norfolk) in the previous year, prey items tended to be larger and were more similar to those provisioned by Sandwich terns 5

8 (Perrow et al., 2010). The larger size meant that, despite having a lower feeding rate, common tern chicks at Saltholme received substantially more energy per hour than those at Blakeney Point in 2008 (Perrow et al., 2010). Gibraltar Point, Lincolnshire A study of a small number of little tern nests at Gibraltar Point, Lincolnshire, in 1978 reported that chicks were predominantly fed on crustaceans, which may have made up 90% of the diet (Davies, 1981). Most identified items were Natantia prawns, with the remainder of the diet largely made up of sandeels (Ammodytes). Young chicks were fed small prey such as prawns and fish around 4 cm long, but by 2 weeks of age they accepted fish as large as those consumed by adults (8 cm long) (Davies, 1981). North Norfolk Between 2006 and 2008, Perrow et al. (2011b) visually tracked foraging common and Sandwich terns at the Greater Wash. Both species were observed to catch fish (including Ammodytes and Hyperoplus sandeels, herring and sprat) and invertebrates, but no invertebrates were observed being fed to chicks (Perrow et al., unpublished; cited by Perrow et al., 2011b; Perrow et al., 2011b). This suggests foraging adults consume low energy prey items themselves and feed their chicks on more energy-rich prey, as has been observed elsewhere (Dunn, 1972; Baird, 1991; Wilson et al., 2004). In 2008 the main prey fed to Sandwich tern and common tern chicks at Blakeney Point were clupeids, with sandeels being of secondary importance for Sandwich terns but rare in the diet of common terns (Perrow et al., 2010). Although Sandwich terns fed their chicks significantly larger prey than common terns, common terns provided more biomass per hour due to a significantly higher provisioning rate (Perrow et al., 2010). East Norfolk Clupeids are an important prey for little terns on the east coast of Norfolk. During observations of little tern chick provisioning in 2002 and 2003 at Winterton, clupeids (herring and sprat) composed >60% of prey items in both years (Perrow et al., 2004). The majority of the remaining prey items went unidentified, while sandeels and invertebrates each contributed 8% of chick diet. Observations at North Denes from 2002 to 2006 also found that clupeids were the most frequent prey items fed to little tern chicks, accounting for an average of 82% of identified items (Perrow et al., 2011). Clupeids, particularly herring, were also the predominant fish caught during trawl samples (96% of fish), with the greater sandeel (H. lanceolatus) being present at low numbers (Perrow et al., 2011). The presence of larval clupeids in the area suggests spawning grounds are nearby, which may have influenced the distribution of little tern colonies along the east coast (Perrow et al., 2004). However, the abundance of clupeids may have been negatively affected by construction of the Scroby Sands wind farm, which was carried out from November 2003 to March Following construction of the wind farm there was a large reduction in young of the year (YOY) herring, with very low abundances in (Perrow et al., 2011). This reduction could not be explained by other environmental factors such as the North Atlantic Oscillation (NAO), sea temperature or water clarity (Perrow et al., 2011). Clupeids continued to dominate the diet of little tern chicks at North Denes in despite the decline in the abundance of YOY herring, but foraging success (number of fish caught per minute) was significantly lower (Perrow et al., 2011). This may have been the driver behind the high nest abandonment and total breeding failure of little terns at North Denes in 2004, and nest abandonment at the nearby Winterton colony in 2004 and 2005 (Allen-Navaro 2006, cited by Perrow et al., 2011; Perrow et al., 2011). 6

9 West Scotland The Firth of Clyde In the Firth of Clyde between 1951 and 1980 there was a decline in breeding numbers of small, surfacefeeding seabirds, specifically Arctic, common, Sandwich and roseate terns and black-legged kittiwakes (Monaghan and Zonfrillo, 1986). In contrast, larger seabird species and those that targeted larger prey, could dive deeper and could travel further were observed to increase or remain stable (Monaghan and Zonfrillo, 1986). This disparity may have been partially caused by surface-feeding species depending on a narrower range of prey (predominantly sandeels and clupeids) than larger, more generalist species, and therefore demonstrating a greater response to changes in prey availability (Pearson, 1968; Monaghan and Zonfrillo, 1986). While changes in the stocks of sandeels and sprat in the Firth of Clyde over this period were unknown, the herring stock off the west coast of Scotland was known to have collapsed in the 1970s due to overfishing (Simmonds, 2001; Dickey-Collas et al., 2010). An index of the abundance of juvenile herring in the Clyde was calculated using fisheries data from 1974 to 1982 and related to the number of breeding Arctic and common terns on Horse Island over the same period (Monaghan and Zonfrillo, 1986). A significant positive relationship was found, with variation in the abundance of juvenile herring explaining half of the variation in numbers of breeding terns (Monaghan and Zonfrillo, 1986). Although other factors such as mortality in the overwintering grounds, habitat changes and competition with gulls are likely to have played major roles in the decline of terns, food availability during the breeding season may have contributed (Monaghan and Zonfrillo, 1986). The Irish Sea Anglesey The island of Anglesey, northwest Wales, is home to a number of tern colonies which feed primarily on sandeels, with clupeids generally being the second most frequent prey (Newton and Crowe, 2000; Roseate Tern LIFE Project unpublished data). Although the colonies are fairly close to one another, tern diet varies between locations. Sandeels (A. marinus, A. tobianus and Hyperoplus lanceolatus) were the main prey of Arctic tern chicks at Ynys Feurig and the Skerries in while clupeids (mostly sprat, some herring) were of secondary importance (Newton and Crowe, 2000). However, while sandeels composed at least 90% of prey fed to Arctic tern chicks at the Skerries in all years of the study, at Ynys Feurig the percentage of sandeels fluctuated between about 60 and 95% (Newton and Crowe, 2000). The proportions of clupeids and other prey items (mostly squid and crustaceans) were therefore substantial in some years at Ynys Feurig, whereas at the Skerries the proportions of these prey groups were consistently low. Access to a long-term data set ( , some years missing) of tern diet provisioning observations from the Skerries and Ynys Feurig has demonstrated that, in accordance with the study by Newton and Crowe (2000), sandeels have generally accounted for a greater proportion of the diet of Arctic tern chicks at the Skerries than Ynys Feurig (RSPB unpublished data). Of the 18 years between 1993 and 2016 where Arctic tern chick diet composition is known for both sites, sandeels composed 15-98% (mean = 69%) of identified items fed to chicks on Ynys Feurig compared with 50-97% (mean = 81%) on the Skerries. Over the same time period, clupeids composed 0-83% (mean = 22%) of items fed to chicks on Ynys Feurig compared with 0-48% (mean = 18%) of items on the Skerries. Clupeids were the most frequent prey fed to Arctic tern chicks at the Skerries in 1989, but since then sandeels have consistently been the main prey at this colony. In contrast, clupeids were the most frequent prey at Ynys Feurig from At both sites the percentage of chick diet formed by clupeids was relatively high in 2009 compared to previous years (Figure 2). Relatively high numbers of clupeids continued to 7

10 be fed to chicks at Ynys Feurig until 2015 when more typical proportions were observed, whereas the pattern at the Skerries was more variable. Clupeids are often more abundant in inshore areas while the lesser sandeel generally associates with offshore areas (Reay, 1970; Heessen et al., 2005, 2015), perhaps explaining the higher proportion of clupeids and lower proportion of sandeels in the diet of terns at Ynys Feurig (inshore of Anglesey) compared to the Skerries (located about 3 km offshore of the northwest coast of Anglesey). In some years provisioning observations were either not carried out or the data was omitted due to quality issues; these years were 1992, 1997, and 2007 at the Skerries and 2008 and 2013 at Ynys Feurig. Figure 2. The percentage of prey items fed to Arctic tern chicks during provisioning observations at the Skerries (red) and Ynys Feurig (blue) that were clupeids (top) and sandeels (bottom), for (RSPB unpublished data). The dashed red line shows the mean percentage over all years at the Skerries; the dotted blue line shows the mean percentage over all years at Ynys Feurig. In 1999 Sandwich terns at Cemlyn Bay fed their chicks predominantly on sandeels, with clupeids (mostly sprat, some herring) contributing the remainder (Newton and Crowe, 2000). In 2009 Sandwich terns continued to feed chicks mostly on sandeels and clupeids, although they also provided a small number of gadoids and rockling (Perrow et al., 2010). Clupeids accounted for a greater proportion of overall energy delivered to Sandwich tern chicks at Cemlyn Bay despite being fed to chicks less frequently than sandeels (Perrow et al., 2010). 8

11 Rockabill Island In contrast to the Anglesey tern colonies on the opposite side of the Irish Sea, clupeids are the main prey of breeding terns at Rockabill. From 1997 to 1999 clupeids were the most frequent prey fed to roseate tern chicks, with sandeels of secondary importance and gadoids fairly infrequent (Newton and Crowe, 2000). Clupeids (mostly sprat) have continued to be the main prey fed to roseate tern chicks in recent years, contributing >70% of roseate tern chick diet in and 2015 (Hulsman et al., 2007; BirdWatch Ireland, 2008, 2009, 2010, 2011; Burke et al., 2016; Cummins et al., 2016). Sandeels and gadoids have generally remained of secondary importance (BirdWatch Ireland, 2010, 2011; Burke et al., 2016). An exceptional year was 2016, when sandeels and clupeids each contributed about 50% of roseate tern chick diet (Burke et al., 2016). Common terns at Rockabill also feed their chicks mainly on clupeids (Hulsman et al., 2007; BirdWatch Ireland, 2008; Burke et al., 2016). However, in 2009 and 2011 gadoids were the predominant prey fed to common tern chicks (BirdWatch Ireland, 2009, 2011) and were often the second most frequent prey in other recent years (Burke et al., 2016). Sandeels were more frequent in 2016, as was observed for roseate terns, but still only contributed 12% of prey fed to common tern chicks (Burke et al., 2016). This suggests that, while both terns primarily feed their chicks on clupeids, common terns are more dependent on gadoids as a secondary prey whereas roseate terns are more dependent on sandeels. Absent from the colony since 2007, snake pipefish were brought to nests at Rockabill from 2014 to 2016 (Burke et al., 2016). The majority were found at common tern nests, suggesting they are targeted more often by this species than by roseate terns (Burke et al., 2016). The reoccurrence of pipefish at the colony may have been caused by increased abundance of the fish in the North Sea, or a reduction in the availability of key prey species, such as sprat (Burke et al., 2016). Lady s Island Lake Lady s Island Lake is a brackish lake situated on the Irish coast between the southern Irish Sea and northern Celtic Sea. Although clupeids were the most important prey at Rockabill in the late 1990s, chick diet at Lady s Island Lake over the same period was more similar to that at Anglesey, with sandeels being the most frequent prey of Arctic, common, roseate and Sandwich terns (Newton and Crowe, 2000). Clupeids (mostly sprat, some herring) formed the majority of the remaining diet, generally contributing <40% of prey items (Newton and Crowe, 2000). The Celtic Sea The Isles of Scilly On the Isles of Scilly observations have been made of common terns carrying food items into colonies. Although this is likely to include both courtship feeding and chick provisioning and is therefore not directly comparable to other studies discussed here, it is apparent that clupeids are rare or absent from the tern diet at these colonies (Isles of Scilly Wildlife Trust, 2005). From the most frequent prey brought to the colonies were sandeels which contributed 80-90% of items, with shrimp being the second most frequent prey group and other prey types contributing just 1-2% of prey (Isles of Scilly Wildlife Trust, 2005). The Southern North Sea The Haringvliet 9

12 Between 2012 and 2015 observations were made of chick provisioning at a Dutch Sandwich tern colony on the Haringvliet, a large inlet in the southern North Sea (Fijn et al., 2016). Clupeids were the main prey, contributing 77% of prey fed to chicks, with sandeels (20%) and other prey (3%) making up the remainder of the diet (Fijn et al., 2016). Combining GPS tracking data of foraging Sandwich terns with observations of chick provisioning indicated that larger sandeels and clupeids were caught on longer trips relatively further offshore, while small clupeids were caught during shorter trips nearer the colony (Fijn et al., 2016). This was supported by a significant positive relationship between average trip length and the length of clupeids and sandeels brought back to the nest (Fijn et al., 2016). Thus, in order to catch large prey items for their chicks, Sandwich terns at this colony must spend more time foraging and travel further from the colony. The Wadden Sea, south-eastern North Sea The Wadden Sea is located in the south-eastern North Sea. As is observed on the Haringvliet, sandeels are less important to seabirds in this region because of a greater abundance of sprat and juvenile herring than in the northern North Sea (Daan et al., 1990; Furness and Tasker, 2000; Harris et al., 2006). Terns at coastal colonies near inland waters in this region have access to a greater variety of prey than terns at island colonies. This was demonstrated by Becker et al. (1987) who collected dropped prey items and pellets at nesting sites of common terns on islands in the Wadden Sea and along the German coast in They found that colonies breeding on islands in the Wadden Sea predominantly fed on clupeids, flatfish and crustaceans, while sticklebacks were rarely caught. In contrast, sticklebacks appeared to be an important prey for terns at coastal colonies and were mainly caught in inland waters. Due to the biases associated with the analysis of dropped prey and pellets (Barrett et al., 2007), clupeids were probably underrepresented while sticklebacks and flatfish were overrepresented. Nevertheless, coastal colonies appeared to have a greater variety of prey items available to them than island colonies due to the availability of freshwater fish (Becker et al., 1987). Dutch Wadden Sea - Griend Sandwich tern chicks on Griend are fed almost entirely on clupeids (herring and sprat) and sandeels (A. tobianus and H. lanceolatus), with the dominant prey varying between years (Veen, 1977; Brenninkmeijer and Stienen, 1994; Stienen et al., 2000). Sandeels tended to be caught in areas with strong currents, while clupeids were targeted in shallow water or when driven to the surface by predatory fish (Veen, 1977). As in other areas, Sandwich terns at this colony feed their chicks on increasingly large fish as the chicks grew older, with the length of clupeids and sandeels ranging from 1.5 cm to 21.5 cm (Veen, 1977; Stienen et al., 2000). However, one study found that compared to other Sandwich tern colonies, the average weight of fish supplied per chick per day was low at Griend, indicating low food availability in the area (Brenninkmeijer and Stienen, 1994). This may have contributed towards the low number of breeding Sandwich terns at Griend in the 1980s and 1990s (Brenninkmeijer and Stienen, 1994). Common terns at Griend also suffer from low availability of high quality prey. In contrast to populations away from the Wadden Sea, the number of common tern pairs on Griend declined between 1994 and 2006, probably due to a combination of emigration and poor breeding success (Stienen et al., 2009). Stienen and Brenninkmeijer (1998) found that common terns on Griend had the lowest breeding success of any colony along the Dutch coast, and deduced that this was probably related to food supply. First of all, other possible drivers of low breeding success such as low habitat availability, high disturbance, pollution or predation were not observed during the study period, although predation has been a problem in other years (Stienen et al., 2009). Secondly, a study by Stienen and van Tienen (1991) showed low 10

13 availability of energy-rich prey around Griend, with almost 40% of the food provisioned to chicks in 1989 and 1990 being low energy prey such as flatfish, shrimp and crabs compared to 10% at other colonies in the Wadden Sea, Germany and around the United Kingdom. Finally, while the main cause of chick mortality in German colonies was flooding or predation, starvation was the predominant driver on Griend (Stienen and Brenninkmeijer, 1998). German Wadden Sea - Banter See and Minsener Oldeoog A study of common terns at Banter See, an artificial brackish lake on the German Wadden Sea coast, supported previous findings that terns from coastal colonies forage both inland and at sea (Becker et al., 1987; 1997). Terns at Banter See were observed to forage in the lake for limnetic prey, such as the three-spined stickleback Gasterosteus aculeatus, and in the sea for marine prey, such as herring (Becker et al., 1997). Although terns at this site have access to multiple prey species, they are still vulnerable to low availability of high-energy food. During a heat wave in 1989, where the average daily water temperature exceeded 20 C, common tern chicks at this colony were observed to be fed predominantly on sticklebacks (80% of prey) and very few herring (3%; Becker et al., 1997). This corresponded with low growth rates and relatively high mortality of common tern chicks >18 days old. In 1990, when temperatures were around 18 C, the main prey fed to chicks was herring with sticklebacks of secondary importance (Becker et al., 1997). In contrast to the previous year, no chicks >18 days old died (Becker et al., 1997). Thus a higher proportion of herring in the diet coincided with increased chick survival. Analysis of fisheries catch data showed a significant negative correlation between the mean water temperature and the mean catch of herring in 1989, with herring being largely absent from the Wadden Sea during the heat wave (Becker et al., 1997). This suggests the low survival of common tern chicks during the heat wave was caused by a temperature-driven loss of the main prey, herring, from the Wadden Sea (Becker et al., 1997). Due to a lack of freshwater prey, terns at island colonies feed their chicks entirely on marine prey. On the island of Minsener Oldeoog in , Frick and Becker (1995) found that common terns fed their chicks on larger prey and more clupeids and sandeels than Arctic terns, which instead fed mostly juvenile fish, shrimp and crabs to their chicks. This reflects a vastly different diet to that of Arctic terns breeding in the UK, where sandeels tend to be the main prey. Due to the larger size of prey and greater proportion of energy-rich fish, common terns fed their chicks three times less frequently than Arctic terns but achieved higher breeding success in both years. Based on the evidence showing clupeids as a major prey of common terns in the Wadden Sea, Danhardt and Becker (2011) investigated the relationship between herring and sprat populations and long-term productivity ( ) of common terns at Minsener Oldeoog and Banter See. From 2002 to 2009, poor breeding success of common terns, including total breeding failure at Minsener Oldeoog, coincided with poor recruitment of herring in the North Sea and low sprat abundance in the Wadden Sea. When analysed in a multiple regression, herring recruitment and sprat abundance explained 75% of the variation in breeding success at Banter See and 65% at Minsener Oldeoog (Danhardt and Becker, 2011). This provides further evidence that energy-rich prey, such as clupeids, are important drivers of variation in tern breeding success. Summary This review demonstrates that tern diets vary between species, colonies and years. However, it is apparent that terns rely heavily on two main prey groups: sandeels and clupeids (sprat and juvenile herring). Sandeels are important prey for all tern species in the UK but are less frequently consumed in the southern North Sea. They are a particularly crucial food source for seabirds on the Shetlands where 11

14 other small, high-energy fish are less abundant, and appear to be the main prey of most Arctic tern colonies in the UK in most years. However, sprat and juvenile herring are also of high importance for breeding terns at most colonies in the UK and Ireland, excluding Shetland and the Isles of Scilly. Indeed, sprat appears to be the most important prey for common terns at the Firth of Forth and common terns and roseate terns at Rockabill Island, which currently supports the largest roseate tern colony in Europe (Burke et al., 2016). The ecology and status of sandeels, sprat and herring Given the importance of sandeels, sprat and juvenile herring for breeding terns in the UK and Ireland, consideration must be given to how populations of these fish are likely to respond to long-term environmental change. This requires a review of the ecology of the three prey species including information relating to their diets, as one of the major manifestations of climate change in the North Sea has been a regime shift in the plankton community (Beaugrand, 2004). Additionally important are the current distributions and habitat requirements, as these will affect the degree to which species are able to shift in response to changes in the marine environment. Recent and current trends are summarised to indicate the status of these species and review whether populations are already responding to environmental change. The importance of these prey species for other seabirds is also discussed, as is the current fishery pressure. Sandeels Sandeels are small pelagic fish in the sand lance family (Ammodytidae). They have a moderately high energy content (Harris and Hislop, 1978) and a long, thin body shape which makes them easy for seabird chicks to swallow. They exhibit diel and seasonal burying behaviour, embedding themselves in the sand during the hours of darkness and throughout most of the winter (Winslade 1974; Freeman et al., 2004; van der Kooij et al., 2008), during which time they become largely inaccessible to seabirds (although see Watanuki et al., 2008). Five sandeel species are found in the North Sea: the lesser sandeel Ammodytes marinus, the small sandeel A. tobianus, the greater sandeels Hyperoplus lanceolatus and H. immaculatus and the smooth sandeel Gymnammodytes semisquamatus. These sandeels differ in their distribution and abundance in the North Sea. The lesser sandeel is the most common of the five species (ICES, 1997; Cabot and Nisbet, 2013) and is predominantly found in offshore waters, whereas its close relative and the second most frequent species A. tobianus typically occurs in inshore areas (Reay, 1970). Due to the importance of the species for avian predators, piscivorous fish, marine mammals and the industrial sandeel fishery in the North Sea (Monaghan, 1992; Pierce & Santos, 2003; Pierce et al., 2004; Sharples et al., 2009), the following information relates primarily to A. marinus, hereafter referred to as sandeels. For a more detailed review of the lesser sandeel, see Green (2017). Diet Sandeels preferentially feed on larger copepods, such as Calanus species, over smaller copepods, such as Temora (van Deurs et al., 2013). Of the Calanus copepods in the North Sea, C. finmarchicus seems to be a particularly important prey species for A. marinus (van Deurs et al., 2009; van Deurs et al., 2013). Sandeels feeding on large C. finmarchicus consume more food (i.e. have a higher stomach content weight) than conspecifics feeding on smaller copepods, possibly due to reduced handling time (van Deurs et al., 2013). Further, C. finmarchicus is critical for the survival of sandeel larvae due to temporal overlap in larval emergence and egg production by this copepod (van Deurs et al., 2009). 12

15 However, the abundance of C. finmarchicus in the North Sea has declined since the mid-1980s and has coincided with an increase in the congeneric C. helgolandicus (Planque and Fromentin, 1996; Beaugrand et al., 2003). While C. finmarchicus egg production peaks in March and abundance peaks in April-June, C. helgolandicus has high egg production but low abundance in May, with peak abundance in July-September (Planque and Fromentin, 1996; Jónasdóttir et al., 2005). Critically, egg production in C. finmarchicus coincides with the hatching period and early life of larval sandeels, whereas that of C. helgolandicus does not (van Deurs et al., 2009). Distribution The lesser sandeel is found in fairly shallow (<100 m) offshore waters in the North Sea (Reay, 1970; Wright et al., 2000) and demonstrates high habitat specificity. Studies show A. marinus has a preference for medium/coarse grained sands ( mm) with a low silt content (Wright et al., 2000; Holland et al., 2005). The greatest abundances of sandeels are found in areas with a bottom temperature of C and a surface salinity of ppt (van der Kooij et al., 2008). This high habitat specificity results in distinct spawning grounds, with notable areas being the Firth of Forth and Moray Firth off eastern Scotland, offshore of the Northumberland coast, Dogger Bank in the central-southern North Sea, offshore of the southeastern coast of England and the eastern Irish Sea (Fisheries Research Services, 2003; Ellis et al., 2012). It is worth noting that although sandeels are a crucial prey for seabirds at Shetland, the habitat in this area is not particularly suitable for sandeels (Jensen et al., 2011; Ellis et al., 2012). Although larval advection is driven by the prevailing sea currents and can therefore vary between years (Wright, 1996; Proctor et al., 1998), sandeel nursery areas broadly match the spawning grounds (Ellis et al., 2012). A further consequence of such narrow habitat and environmental preferences, combined with a limited dispersal propensity of adults, is that sandeels are unlikely to be able to relocate in response to changing conditions, and are therefore considered particularly vulnerable to the effects of climate change and localised fisheries depletion (Engelhard et al., 2008; Heath et al., 2012). Trends An increase in the sandeel total stock biomass from 1974 to 1984 was followed by a rapid decline between 1984 and 1987 (Anon., 1989). Around Shetland there was a marked decline in the total abundance of 0-group sandeels from 1982 to 1988 and a decrease in total production from 1983 to 1987 (Bailey et al., 1991). This was accompanied by a shift in many species of seabird from a predominantly sandeel-based diet to greater consumption of alternative prey, with concurrent declines in several populations (Bailey et al., 1991). A more drastic decline has occurred since the early 2000s, with catches in the North Sea in recent years being around 50% of catches in (Figure 3), and are due largely to a decline in landings from the Dogger Bank area (ICES, 2013b; ICES, 2017a). There has also been a shift in the quality of sandeels as prey. Between 1973 and 2006 the mean lengthat-date of 0-group A. marinus on the Wee Bankie aggregation declined by 22%, corresponding to a 60% decline in energy content (Wanless et al., 2004; Frederiksen et al., 2011). A study in the Dogger Bank area also showed a reduction in mean length-at-date of 0-group and 1-group lesser sandeels over a similar period (van Deurs et al., 2013). Although initially driven by later hatching dates, more recent reductions in length-at-date are suggested to have been driven by lower growth rates and changes in size-dependent mortality (Frederiksen et al., 2011). These changes may be a consequence of climaterelated shifts in the distributions of copepod prey and other planktivorous fish that compete with A. marinus (Frederiksen et al., 2011; van Deurs et al., 2013). For example, van Deurs et al. (2013) observed that the increase in sandeel length-at-date before 1987 and the subsequent decline corresponded with 13

16 an increase and then decline in the abundance of C. finmarchicus before and after the mid-1980s (Planque and Fromentin, 1996). Additional evidence of changes to sandeel populations in the North Sea comes from shifts in seabird diet. Comparing recent diet data to that collected a few of decades earlier suggests there has been a general decline in the proportion of sandeels in the diet of Guillemot chicks at colonies bordering the North Sea (Anderson et al., 2014). The proportion of lesser sandeels in the diet of Common Guillemots chicks at Sumburgh Head, Shetland, declined from 80% to 55% between the early 1990s and late 2000s, with gadoids making up most of the remainder of diet (Heubeck, 2009). A similar dietary shift occurred in the diet of guillemots at Fair Isle between 1999 and This shift coincided with a general reduction in breeding success and fledging weight at Shetland (Heubeck, 2009). Figure 3. Total catches (tonnes) of sandeels in the North Sea from Reprinted with permission from ICES (2017a). Importance for other seabirds Seabirds consume an estimated 200,000 tonnes of sandeels every year in the North Sea alone (Furness and Tasker, 1997). Sandeels make up a major component of chick diet for many seabirds in this region (Harris and Wanless, 1991; Wilson et al., 2004; Harris et al., 2005), and the distributions of some seabirds have been shown to correlate with areas of high sandeel availability (Wright and Begg, 1997; Wanless et al., 1998). However, the importance of sandeels for seabirds varies geographically. The internationally important seabird colonies at Shetland are particularly dependent on sandeels during the breeding season due to a lack of other small, lipid-rich prey (Tasker et al., 1987; Bailey et al., 1991; Furness and Tasker, 1997). A study of the diet of guillemots around the UK from found that, although sandeels were the most common prey fed to chicks, they contributed a higher proportion of the diet on the west coast relative to the east (Anderson et al., 2014). Sandeels are less important to 14

17 seabirds in the southern North Sea due to an abundance of juvenile sprat and herring (Furness and Tasker, 2000). Seabirds that feed on prey in surface waters, such as kittiwakes and terns, and those that rely on a small number of prey species, such as colonies on Shetland and Orkney, are particularly sensitive to local depletion of sandeels (Furness and Tasker, 2000; Daunt et al., 2008; Frederiksen et al., 2005; Frederiksen et al., 2008; Eliasen, 2013). Indeed, during operation of a local, inshore sandeel fishery near the Shetlands, a decline in landings since the mid-1980s coincided with large declines in the breeding success of surface-feeding seabirds, particularly Arctic terns and black-legged kittiwakes (Bailey et al., 1991; Monaghan, 1992; Furness and Tasker, 1997). Carroll et al. (2017) found kittiwake breeding success at a colony on the Yorkshire coast of England was positively associated with high sandeel SSB in the preceding winter, but negatively correlated with sandeel fishing mortality at Dogger Bank two years earlier, suggesting that fishing intensity may (by a mechanism not yet known) be having a negative effect on kittiwake productivity. A long-term study of seabirds at Foula, Shetland, showed that the annual estimated total stock biomass of sandeels in the Shetland area could explain annual variation in the breeding success of surface-feeding seabirds moderately well (Furness, 2007). A decline in sandeel availability at Shetland correlated with a 75% reduction in breeding success of Great skuas over three years (Klomp and Furness, 1992), with a longer study finding a similar pattern (Oswald et al., 2008). Additionally, declines in Arctic skua breeding success and chick growth rates at Shetland have been shown to coincide with declines in local sandeel availability (Caldow and Furness, 1991; Phillips et al., 1996). Temporal overlap between sandeel peak availability and seabird chick rearing is also a key driver of breeding success (Harris and Wanless, 1991; Wright, 1996; Rindorf et al., 2000). For example, between 1983 and 2006 the mean date at which 0-group sandeels reached a threshold size became significantly later, with an average delay of almost two weeks per decade (Burthe et al., 2012). Over the same period, the timing of seabird chick rearing in some species on the Isle of May became later (Burthe et al., 2012). However, the phenological shift in chick rearing lagged behind shifts in 0-group size, resulting in a net decline in the length of 0-group sandeels being fed to chicks (Burthe et al., 2012). Fisheries The North Sea sandeel fishery began in the late 1950s, following collapse of the heavily fished herring and mackerel stocks, and is now the largest single-species industrial fishery in the region (Furness, 2003). Initially fished at a low level, by 1989 sandeel catch exceeded 1 million tonnes (Furness, 2003) and between 1983 and 2002 the average annual landing was >800,000 tonnes (ICES 2012). However, following a large peak in the late 1990s the annual landings declined and between 2003 and 2010 the average annual catch was 313,000 tonnes (ICES 2012). If stocks of piscivorous fish species that predate on sandeels recovered, current fishing levels would be likely to have a negative impact on large populations of seabirds (Furness, 2002). Indeed, mackerel, cod and other fish predators of sandeels have recovered to a significant extent in recent years and the trend is ongoing, with improved fisheries management a contributory factor (European Commission, 2017). During early operation of the North Sea sandeel fishery it was suggested the fishery had no impact on breeding seabirds (Furness, 2003). However, concerns about the potential impacts of fishing sandeels on seabird breeding success have occasionally led to temporary or long-term closures of fisheries (Greenstreet et al., 2010). A small, inshore sandeel fishery at the Shetland Isles opened in 1974, with annual landings peaking at just 52,000 tonnes (Dunn, 1998). Despite its small size, the fishery overlapped substantially with foraging areas of internationally important seabird colonies breeding at 15

18 Shetland (Dunn, 1998). Collapse of the Shetland fishery and the subsequent decline in seabird breeding success around the Shetland Isles in the 1980s led to the precautionary closure of the fishery in 1990 (Hamer et al., 1993; Dunn, 1998; Rindorf et al., 2000; ICES, 2005; Greenstreet et al., 2010). However, although the sandeel population declined during fishery operation and recovered following its closure, this may have been caused by variation in 0-group immigration and survival rather than changes in fishery pressure (Wright, 1996). The Shetland fishery re-opened in 1995 under restrictions of a TAC of 7,000 tonnes and closure during the months of June and July to prevent competition with seabird colonies (ICES, 2005) but is no longer operational due to the severe and chronic reduction of sandeels from Shetland waters over the past decade (E Dunn, pers. comm). A section of the larger, Danish-led industrial sandeel fishery in the North Sea was closed permanently in 2000 due to concerns regarding the negative impact the fishery was having on sandeel-dependent seabirds, particularly kittiwakes (Camphuysen 2005; Daunt et al., 2008; Frederiksen et al., 2008). The closure area, which remains in place today, includes the Wee Bankie, Marr Bank and Berwick Bank and extends from northeast Scotland to Northumberland, covering an area of 20,000 km 2 (De Santo and Jones, 2007; Daunt et al., 2008). The closure appears to have caused an immediate recovery of sandeels within the area, with an observed increase in the biomass of 0-group and 1+ sandeels on the Wee Bankie within the first year (Greenstreet et al., 2006). However, by 2007 the spawning stock biomass (SSB) had declined to levels similar to those when the fishery was operating (Greenstreet et al., 2006; Greenstreet et al., 2010). This decline indicated that, even without fishing mortality, sandeel abundance will decline if recruitment cannot compensate for natural mortality (Greenstreet et al., 2006; Greenstreet et al., 2010). Thus, in addition to monitoring the abundance of sandeels while a fishery is active, it is also important to monitor sandeel abundance after a fishery closes in order to assess the effectiveness of the closure (Greenstreet et al., 2010). Sprat The European sprat, Sprattus sprattus, is a small pelagic fish in the family Clupeidae (Heessen et al., 2005). Sprat may be particularly important prey for breeding terns due to their high energy content. Sprat have a higher calorific value per gram than sandeels, with one study reporting that sprat provided an average 10.9 kj -g while sandeels provided an average 6.5 kj -g (Harris and Hislop, 1978). This pattern was observed in a later study, which reported very low energy content of sandeels <10 cm long (Hislop et al., 1991). Studies of the energetic content of herring have found that while large mature herring have a high calorific value, small herring of a suitable size for most seabirds have a low energy content, similar to that of gadoids (Barrett et al., 1987; Hislop et al., 1991). Thus, of the three prey species considered here, sprat appears to be the most energetically favourable. Diet Analysis of stomach contents from across the species geographical range shows that sprat feeds predominantly on crustaceans, such as copepods, euphausiids, amphipods and decapods (Kleinertz et al., 2012). Calanoid copepods contributed the most to the stomach weight of sprat in the North Sea, while harpacticoid copepods contributed the most in the English Channel (Kleinertz et al., 2012). In the inshore waters around western Scotland, sprat predominantly target copepods, including Temora longicornis, C. finmarchicus and Centropages typicus (De Silva, 1973). T. longicornis is also an important prey in the southern Baltic Sea during the autumn (Casini et al., 2004). This is supported by the observation that an increase in the abundance of T. longicornis in the Baltic Sea in the late 1980s, as part of a temperature-driven regime shift in the marine environment, coincided with an increased abundance of sprat (Alheit et al., 2005). A concurrent regime shift occurred in the North Sea, with an 16

19 increase in C. helgolandicus but a decline in C. finmarchicus (Beaugrand, 2003). C. helgolandicus is a warm-temperate species (Planque and Fromentin, 1996) and T. longicornis is eurythermic, being able to withstand and adapt to a wide range of temperatures (Moison et al., 2012). Analysis of long-term data showed both copepod species have undergone an overall increase in abundance around the UK and Ireland since the 1950s (Pitois and Fox, 2006). There is evidence to suggest C. helgolandicus is also an important prey for sprat in the North Sea (Fauchald et al., 2011; Lynam et al., 2017). Thus, the increases in C. helgolandicus and T. longicornis may have befitted sprat populations. In contrast, the large, coldwater species C. finmarchicus (van Deurs et al., 2009; van Deurs et al., 2013), a major prey of sandeels and herring, has shown a decline in this region (Pitois and Fox, 2006). Distribution Sprat is found in relatively shallow waters ( m deep), inshore areas and estuaries in the eastern Atlantic (Whitehead, 1985; Daan et al., 1990; Araujo et al. 2000; Nedreaas et al., 2015). The greatest abundances are found at depths of m in the Celtic Sea and m in the North Sea (Heessen et al., 2015). At the northern limits of its distribution, sprat is found in the Irish Sea, North Sea and southeastern Norwegian Sea (Nedreaas et al., 2015). In the North Sea sprat is abundant south of Dogger Bank but much less so further north, being patchily distributed along the northeast coast of Britain and scarce around Shetland (Daan et al., 1990; Brown and Pierce, 1997; Furness and Tasker 2000; Heessen et al., 2005; Macdonald and Napier, 2014). Geographically isolated populations exist in the Moray Firth and Firth of Forth in north-eastern and eastern Scotland, respectively (ICES, 2013a). These populations are likely to behave as distinct stocks due to limited connectivity with the main sprat stock in the southern North Sea (ICES, 2013a). Sprat populations in the Moray Firth and Firth of Forth are important for a range of natural predators, including seabirds. However, both stocks are vulnerable to local depletion due to isolation from the main sprat stock, and past declines in local sprat abundance have been linked with declines in predatory birds in these areas (Jennings et al., 2012; ICES, 2013a). There is a substantial population of sprat in Lyme Bay, the most heavily fished area of the English Channel, with smaller numbers offshore and further west (ICES, 2013a). Large numbers also exist in the eastern English Channel (ICES, 2013a). In the Bristol Channel large schools of mostly juvenile sprat have been observed in recent years (ICES, 2017a); the amount of movement between this population and those in the English Channel is currently unknown (ICES, 2013a). In the Celtic Sea and west of Scotland the structure of the sprat stock is unclear (ICES, 2013a; ICES, 2017a) and there is little information relating to the abundance or distribution of sprat in this area. However, there is some evidence to suggest that sprat spawned in the Irish Sea are transported both northwards, into areas such as the Firth of Clyde, and southwards, into the Celtic Sea (ICES, 2013a). Further, a substantial abundance of sprat in the Irish Sea is indicated by the relatively high occurrence of this species in the diet of seabirds, compared to other regions (Newton and Crowe, 2000; Hulsman et al., 2007; Chivers et al., 2012; Cummins et al., 2016). Mature sprat tend to spawn in offshore areas where they form large aggregations (ICES, 2013a). Spawning occurs in the spring and summer, with females spawning multiple times throughout the season (Alheit, 1988; Nedreaas et al., 2015), and around much of the UK excluding the Shetland Islands. Concentrations of both larval and mature sprat off the west coast of Orkney suggests this is a spawning area (Hopkins, 1986), although the presence of larvae varies between years (Mackay, 1984). This may be caused by variation in oceanic currents. As sprat are pelagic spawners their eggs and larvae are dispersed in the open sea (Hopkins, 1986); thus the distribution of larval sprat is determined largely by the prevailing sea currents (Baumann et al., 2006). This can result in considerable mixing of larvae from different spawning populations (ICES, 2013a). For example, eggs and larvae that are spawned in the 17

20 English Channel drift eastwards into the southern North Sea (ICES, 2013a). Under certain circulations, individuals from this spawning area coalesce with other spawning populations in a nursery area in the German Bight (Corten, 1986; Baumann et al., 2009). Variation in sprat distribution can have serious consequences for seabirds; for example, a large-scale shift in the distribution of sprat in the late 1970s early 1980s may have contributed to increased mortality rates of auks in the northwest North Sea (Blake, 1984; Harris and Bailey, 1992). Trends While sandeels are being negatively affected by climate change in the North Sea, sprat appear to have increased around the UK since the 1990s, although there is a lack of quantitative data regarding stocks in this region (Rijnsdorp et al. 2010; ICES, 2012; Heessen et al., 2015; Nedreaas et al., 2015). Localised depletion of stocks of the geographically isolated population in the Firth of Forth occurred in the 1980s due to targeted fishing (Heath et al., 2012; Jennings et al., 2012). However, following the removal of fishing pressure and an improvement in climatic conditions (a shift from a cold-water to warm-water regime; Beaugrand et al., 2002), sprat stocks appear to have recovered and are once again supporting a large population of common terns in this area (Heath et al., 2012; Jennings et al., 2012; ICES, 2013a; Anderson et al, 2014). An acoustic survey indicated the abundance and biomass of sprat in the North Sea more than doubled between 2015 and 2016, exceeding 1,100,000 t (ICES, 2017a). Recruitment in the North Sea in 2016 is estimated to be the highest on record (ICES, 2017b). Further, an index of the abundance of 1-year old sprat in 2017, based on International Bottom Trawl Survey (IBTS) estimates, was the highest since records began in 1974 (ICES, 2017a). The acoustic survey indicates that the last three years of the period 2004 to 2016 were the years of highest sprat biomass in the North Sea (ICES, 2017a). While the North Sea sprat stock appears to have increased in recent years, acoustic surveys in the English Channel indicate there was a sharp drop in biomass in compared to previous estimates and a decline in landings per unit effort (LPUE) in 2015 (ICES, 2017a). Importance for other seabirds As an abundant forage fish, sprat provides an important intermediate link between lower and higher trophic levels (Cardinale et al., 2002; ICES 2011). Sprat migrate daily within the water column, being more abundant in surface waters at dawn and dusk and forming aggregations at the sea bottom during the day (Cardinale et al., 2003; Nilsson et al., 2003; Falkenhaug and Dalpadado, 2014). This behaviour is likely in response to diel migration of their zooplankton prey in the water column (Williams and Conway, 1984; Cardinale et al., 2003). Thus sprat are more accessible to surface-feeding seabirds during the morning and evening. The importance of sprat for tern chicks was discussed earlier. However, this fish is also an important prey for terns during courtship feeding (Taylor, 1979) and other seabirds breeding in the North Sea. Indeed, fluctuations in the biomass of sprat in the North Sea have had severe consequences for populations of seabirds, such as auks (Blake, 1984). Between the late 1970s and early 1980s the biomass of sprat in the northwest North Sea underwent a large decline, likely contributing to the observed mass starvation of auks, particularly razorbills, on the North Sea coasts of England and Scotland (Blake, 1984; Harris and Bailey, 1992). This may have been caused by a largescale south-easterly shift in the distribution of sprat from the northwest North Sea (Harris and Bailey, 1992). Winter mortality rates of puffins and guillemots on the Isle of May during this period were also linked to the abundance of sprat, which may be particularly important for seabirds during the winter because of the tendency of sandeels to remain buried in the sand (Harris and Bailey, 1992; van der Kooij et al., 2008). 18

21 While lesser sandeels are the main prey of black-legged kittiwakes around much of the UK (Lewis et al., 2001; Frederiksen et al., 2005; Harris et al., 2005; Furness, 2007), sprat are an important prey for this seabird in some areas. At the Isle of May, sprat has become increasingly common in the diet of kittiwake chicks and has been the predominant prey since 2000 (Anderson et al., 2014). In the Irish Sea, a study by Chivers et al. (2012) covered just two sites over two years but found some evidence to suggest there may be a link between kittiwake breeding success and the proportion of sprat in the diet. To investigate this further, correlations between kittiwake breeding success and sprat abundance were examined at a regional scale (ICES, 2013a). The years of lowest kittiwake breeding success in northwest Scotland did overlap with the years of lowest sprat catch on the west coast of Scotland, but the contribution of sprat to kittiwake diet in this area is not known. Kittiwake breeding success in northeast Ireland and southwest Scotland was weakly correlated to sprat stock biomass in the Irish Sea, as estimated from acoustic surveys in the previous autumn (ICES, 2013a). No such relationship was found for colonies on the Isle of Man, northwest England or in the Firth of Clyde (ICES, 2013a). Thus while the abundance of sandeels is an important driver of kittiwake breeding success in some areas (Harris and Wanless, 1997; Lewis et al., 2001; Daunt et al., 2008; Eerkes-Medrano et al., 2017), this may also be true for sprat. The availability of sprat to foraging seabirds is likely to be influenced by the habitat in the areas surrounding the colonies. From , kittiwakes breeding at estuarine colonies in East Scotland consumed a greater frequency of clupeids (believed to be mostly sprat) than those breeding at marine colonies, which fed predominantly on sandeels (Bull et al., 2004). Sprat are tolerant of a wide range of salinities and are found in estuarine habitat around the UK (Araujo et al., 2000; ICES, 2013a), whereas lesser sandeels are predominantly found offshore (Reay, 1970; Wright et al., 2000). Despite the difference in diet composition, there was no clear distinction between kittiwake breeding success at the marine and estuarine colonies (Bull et al., 2004). This suggests that sprat and sandeels are similarly valuable prey for kittiwakes during the breeding season. Abundance is not the only sprat-related variable affecting seabird populations. A lower body condition of sprat, such as a reduction in weight-at-age, can have negative impacts for seabirds even when the total abundance of sprat is high (Österbrom et al. 2006). In 2004 guillemots on the Isle of May experienced very low breeding success, coincident with low sandeel availability, a high proportion of sprat in chick diet, and a lower energy content of sprat and sandeels than expected (Wanless et al., 2005). In the Baltic Sea, a decline in mean weight of age 4 sprat in the 1990s coincided with a decline in the fledging mass of common guillemot chicks (Österbrom et al. 2006). However, the sprat stock subsequently declined as fishing mortality increased, coinciding with an increase in body condition and energy content of sprat and an increase in common guillemot chick fledging mass (Österbrom et al. 2006). A negative relationship between sprat abundance and body condition suggests that body condition is density-dependent, at least in some areas of the species range. For example, inter-annual variation in the condition of sprat and herring in the Baltic Sea is strongly inversely related to total clupeid abundance (Cardinale et al., 2002; Casini et al 2006). Lower body condition of clupeids at higher abundances is likely to be driven by competition for food, as body condition has been shown to be positively related to weight of zooplankton in the stomach (Casini et al 2006). Poor body condition of prey is particularly detrimental for single-prey loading species such as guillemots and terns, as a reduction in energy content of individual fish causes a reduction in energy per chick feed (Frederiksen et al., 2006; Österbrom et al. 2006; 2008). Sprat fisheries For a more detailed account of sprat fisheries around the UK and Ireland, see ICES (2017a). 19

22 Sprat stocks around the UK and Ireland are treated as three management units, based on the distribution of historic fisheries: the North Sea, English Channel and Celtic Seas (ICES, 2013a). However, it is unclear whether populations within these areas constitute a single stock and how much mixing occurs between units (ICES, 2013a). The majority of sprat caught in the North Sea is reduced to fishmeal and fish oil, although some smaller, targeted fisheries catch sprat for human consumption (Heessen et al., 2005). Between 1996 and 2015 the estimated total catches of sprat in the North Sea fluctuated between a low of 61,100 t in 2008 and a high of 290,400 t in 2015 (ICES, 2016). The vast majority of North Sea catches of sprat are taken by the Danish fishery, which accounted for 82-92% of catches between 2013 and 2016 (ICES, 2016). Over the same four years the annual catch by the UK fishery was less than 50 tonnes. The spatial distribution of sprat catches varies between years, but tends to centre on the southern or south-eastern North Sea (Figure 4; ICES, 2016). In the English Channel most of the catch is taken in the Lyme Bay area (ICES, 2017a). The TAC for this area is substantially lower than that in the North Sea, being just 4,120 t in 2017, but over the past 20 years the quota has not been reached (ICES, 2017a). The annual landings in 2015 were 16% of estimated sprat biomass in the Lyme Bay area, and 6% of the biomass in the wider area (ICES, 2017a). However, the estimated biomass of sprat in the English Channel declined from ~60,000 t in 2015 to ~9,300 t in 2016, and because a harvest rate similar to the previous year was applied, the landings in 2016 accounted for a huge 44% of estimated total sprat biomass (ICES, 2017a). Sprat are a short-lived species and the biomass of stocks is highly variable; therefore the large decline in the estimated biomass between 2015 and 2016 may have been caused by variation in recruitment (ICES, 2017a). However, taking such a high proportion of the stock biomass in years of naturally low populations is a concern, as this may hinder stock recovery. There is no TAC in the Celtic Sea or west of Scotland (ICES, 2017a). Landings were highest off the west of Scotland throughout the 1970s, averaging 7,000 t per year (ICES, 2017a). However after 1978 the landings declined and only increased again in the mid- to late-1990s when they averaged 4,600 t per year (ICES, 2017a). Little fishing took place from but landings have since increased, reaching ~2,200 t in 2016 (ICES, 2017a). In the adjacent Irish Sea, landings were also highest in the 1970s but then dropped when the fishmeal factory closed in 1979 (ICES, 2017a). Excluding a peak in 2003, landings in the southern Irish Sea in the 2000s were <500 t per year, but have since fluctuated between a high of ~7,000 t in 2012 and a low of 16 t in 2014 (ICES, 2017a). Landings in the northern Irish Sea have been consistently <800 t since the late 1980s (ICES, 2017a). However, the accuracy of the landings data is unknown and some years may be overestimated. Landings of sprat to the west of Ireland are low, with no landings reported in 2016, although landings in this area may be underestimated (ICES, 2017a). Annual landings in the Celtic Sea have fluctuated widely since the 1990s between <100 t to >4,000 t. In the past 10 years landings have been relatively high, averaging ~2500 t. Sprat stocks in the North Sea are considered to be within safe biological limits and therefore can be caught as part of the quotas of other species, under a regulation within the EU landing obligation that allows inter-species quota transfers (ICES, 2017b). Such transfers, which allow up to 9% of the quota of the target species to be transferred to non-target species, could put sprat at risk of being overfished in the North Sea as these are not considered in current catch advice (ICES, 2017b). 20

23 Figure 4. Annual spatial distribution of sprat catches in the North Sea for Reprinted with permission from ICES (2017a). Herring The Atlantic herring Clupea harengus is a medium-sized pelagic fish in the family Clupeidae (Heessen et al., 2005). Many tern diet studies that report the importance of clupeids and differentiate between sprat and herring note that sprat tends to be the more frequent component, based on the collection of dropped prey items, local knowledge or other information (Langham, 1968; Newton and Crowe, 2000; Robinson et al., 2001; Harris et al., 2005; Burke et al., 2016). However, it is difficult to differentiate between sprat and juvenile herring, leading to misidentification of the two species during observations and in catch data (ICES, 2013a). This is additionally problematic because sprat and juvenile herring often exist in mixed shoals (ICES, 2013a). Some studies only report the presence of clupeids and do not comment on the relative proportions of herring and sprat (Pearson, 1968; Veen, 1977; Monaghan et al., 1989; Brenninkmeijer and Stienen, 1994; Robinson and Hamer, 2000; Morris, 2012). Yet other 21

24 studies do note the importance of herring in particular (Taylor, 1979; Becker et al., 1997; Perrow et al., 2011). This suggests that juvenile herring may be an important source of food for terns in some areas. Diet Herring preferentially feed on larger copepods such as C. finmarchicus, which appears to be their main prey in the North Sea and Norwegian Sea (Parrish and Saville, 1965, cited by Maravelias et al., 2000; Dalpadado et al., 1996), although Temora species may also be important (Last, 1987; Last, 1989; Raab et al., 2012). Amphipods, euphausiids, fish eggs and post-larval sandeels and clupeids are also consumed fairly frequently, particularly when copepods are not available (Last, 1989; Seger et al., 2007). Juvenile herring in the German Bight demonstrated a less diverse diet than sprat and anchovy, specialising predominantly on Calanus and Temora copepods (Raab et al., 2012). Based on the importance of C. finmarchicus for herring in the North Sea, ongoing long-term declines in the abundance of this copepod are likely to have a negative effect on herring stocks (Planque and Fromentin, 1996; Reid et al., 2003). Indeed, herring is already responding to changes in the North Sea plankton community. A decline in the abundance of C. finmarchicus in response to increasing temperatures has been related to a northwards shift in herring catches, with herring leaving certain areas during extreme heat events (Corten, 2001). Such heat-driven shifts can have severe consequences for breeding terns (Becker et al., 1997). Distribution Herring is close to the southern limits of its distribution in the North Sea (Corten, 2001). The greatest density of adult herring exists to the north and east of Scotland whereas juvenile herring, like sprat, are more abundant in the southern and central-eastern North Sea (Daan et al., 1990; Harris et al., 2006; ICES, 2017a). Although herring migrate southwards past Shetland in the late summer, these fish tend to be >20 cm long (Brown and Pierce, 1997, 1998) and are therefore too large and available too late to be of importance for breeding terns in the area. There is a large herring spawning ground around Orkney and extending to south-western Shetland (Corten, 1986; Ellis et al., 2012), but the larvae are largely carried southwards to the east coasts of Scotland and England and into the central and southern North Sea by currents originating from the North Atlantic (Corten, 1986). Thus juvenile herring appear to be less abundant around Shetland than other areas further south (Daan et al., 1990; Furness and Tasker, 2000; ICES, 2017a). Unlike sprat, herring are demersal spawners and attach their eggs to coarsegrained, gravelly substrates on the seabed, resulting in distinct spawning areas (Hopkins, 1986; Ellis et al., 2012). The timing of spawning varies around the UK, but most herring in the North Sea are autumnspawners (Dickey-Collas et al., 2010). Large spawning areas exist around the eastern Scottish coast, off the Northumberland coast, offshore of East England, in the eastern English Channel and in the eastern Irish Sea (Ellis et al., 2012). This largely corresponds with herring nursery grounds, although these include areas in the western Irish Sea and west of Scotland (Ellis et al., 2012). Trends The mean weight-at-age of autumn spawning herring stocks in the Celtic Sea, Irish Sea and North Sea has declined since 1980 (ICES, 2017a). However, the estimated number of mature autumn spawning herring in the North Sea has increased since 1986, with the two years of highest abundance being 2014 and 2016 (ICES, 2017a). Despite a moderate spawning stock biomass, recruitment of autumn spawning herring in the North Sea has been low since 2002, resulting in low stock productivity (Payne et al., 2009; ICES, 2017a). Low recruitment has been linked to an observed reduction in the survival rates of herring larvae since the 1990s, with very low survival since 2000 (Payne et al., 2009, 2013). It is likely 22

25 this is related to changes in the zooplankton community (Beaugrand, 2003; Payne et al., 2009), specifically a decline in the abundance of C. finmarchicus which, together with warming winter temperatures, has been implicated in a long-term northwards shift in catches of herring in the North Sea (Corten, 2001). Importance for other seabirds In comparison to sandeels and sprat, there is less evidence of herring being an important prey for breeding seabirds around the UK and Ireland. This may be a consequence of the difficulty in distinguishing between juvenile herring and sprat (ICES, 2013a). However, there is an additional consideration to be made. The relationship between herring and seabirds is likely to be more complex than that observed for sandeels and sprat because herring predate on these prey species (Last, 1987, 1989; Frederiksen et al., 2007). Therefore, while herring acts as a food source in some cases, it can also behave as a competitor. Such a relationship may have been detected in the Celtic Sea, where a negative relationship between herring abundance and razorbill breeding success suggests predation of herring on other pelagic prey such as sandeels (Lauria et al., 2012). Similarly, in the south-western Barents Sea, herring predation on capelin may have had a negative effect on the numbers of black-legged kittiwakes (Barrett, 2007). Multiple studies in the North Sea also present evidence to suggest a negative relationship between sandeel stocks and the abundance of herring (Sherman et al., 1981; Furness, 2004; Macdonald et al., 2015). Thus, relationships between seabirds and herring are likely to be more complicated and variable than those with sandeels and sprat. Clupeids are the main prey for breeding kittiwakes at estuarine colonies along the east coast of Britain, but are less important at marine colonies where sandeels predominate in the diet (Bull et al., 2004). At the Isle of May, herring play a minor role in the diet of kittiwake chicks which are fed predominantly on sandeels or, since 2000, sprat (Harris and Wanless, 1997; Anderson et al., 2014). Most research demonstrating the importance of herring for breeding seabirds in the North Atlantic comes from areas away from the UK and Ireland, such as the Norwegian coast. Atlantic puffins at Røst in north-western Norway are highly dependent on the availability of herring during the breeding season (Anker-Nilssen et al., 1997; Durant et al., 2003). Low stocks of herring over a 19 year period coincided with very low breeding success of puffins, whereas the subsequent recovery of herring stocks coincided with an immediate improvement in breeding success (Anker-Nilssen et al., 1997). The importance of herring for seabirds at Røst was further supported by significant positive relationships between the abundance of age-0 herring in the Barents Sea and the breeding success of both puffins and kittiwakes (Anker- Nilssen et al., 1997). However, as mentioned above, the impact of herring on seabirds can be complex. For example, while herring appear to have a positive impact on kittiwakes at Røst, at Hornøy in north Norway the proportion of age-1 herring in kittiwake diets has been shown to have a negative impact on breeding success, possibly due to a negative relationship between herring and capelin (Anker-Nilssen et al., 1997). It should be noted that the spring-spawning Norwegian herring and Barents Sea herring stocks included in these studies are distinct from the largely autumn-spawning stocks of herring found in the North Sea (Dickey-Collas et al., 2010). Herring appears to be an important prey for wintering seabirds in the North Sea. Fauchald et al., (2011) related the winter abundance of 10 pelagic seabirds (including auks, gulls, northern gannets and northern fulmars) to the abundance of herring in the previous year and found a significant positive relationship. Similarly, in the Skagerrak-Kattegat in the eastern North Sea, spatial variation in the winter abundance of immature herring was significantly positively correlated with that of kittiwakes, common guillemots and razorbills (Skov et al., 2000). 23

26 Fisheries Herring is caught by both targeted fisheries and as by-catch. Most of the catch is sold for human consumption, although a small portion is reduced to fishmeal and oil (Heessen et al., 2005). The intensity of fishing for herring increased throughout the 20 th century and the North Sea stock declined from more than 4,000,000 t in the first half of the 1960s to <1,000,000 t in 1974, reaching very low levels by 1977 (Simmonds, 2001). This led to closure of the North Sea fishery and diversion of the fishing effort to the west coast of Scotland, but by the late 1970s this stock had also collapsed (Simmonds, 2001). Both stocks have since recovered (although not to their former status), but while the west of Scotland stock has remained relatively stable, the North Sea fishery again reached unsustainable levels and a recovery plan (halving of TAC) was implemented in 1996 (Simmonds, 2001; Heessen et al., 2005). In the North Sea peak landings are between October and March, although fishing occurs throughout the year (Heessen et al., 2005). Landings in the autumn come largely from the Northern Isles, the Aberdeenshire coast, coastal waters around eastern England and northwest of Dogger Bank, while in the spring the main areas are coastal waters off Lincolnshire and East Anglia (Heessen et al., 2005). During the summer a smaller fishery operates mainly around the Northern Isles and the western central North Sea (Heessen et al., 2005). Herring closure boxes exist off the coasts of Northumberland and North Yorkshire from mid-august to the end of September in order to protect juvenile herring (Anon, 1998). Sprat closure boxes exist in the Moray Firth, Firth of Forth and off Jutland with the purpose of reducing the by-catch of juvenile herring (Commission of the European Communities, 2007). Although the mean total landings of herring in the North Sea have increased in recent years, the SSB is expected to decline between 2017 and 2019 due to the weak year classes in 2014 and 2016 (ICES, 2017a). Off the coasts of western Scotland and northern and western Ireland, fishing mortality is low but SSB has gradually declined since 2002 and recruitment has been very low since 2012 (ICES, 2017a). This is reflected by the landings in this area which have not exceeded 50,000 t since 2002; between 2007 and 2016 the mean total landings were ~26,600 t (ICES, 2017a). Herring boxes previously existed off the north-western coast of Scotland and southern coast of Ireland (Anon, 1998; Commission of the European Communities, 2007); however, these have since been deemed no longer necessary for sustainable exploitation and therefore reopened (Commission of the European Communities, 2007; Anon, 2013). In the Irish Sea there are juvenile closures all year along parts of the east coast of Ireland and the west coast of Britain, in addition to spawning closures along the eastern coast of the Isle of Man during the autumn (ICES, 2017a). In contrast to other areas, SSB and recruitment in the Irish Sea has increased since the mid-2000s. Since 2005 the annual total landings have been fairly stable, reaching around 5,000 t (ICES, 2017a). This is, however, considerably lower than landings in the 1970s, which peaked at around 38,000 t (ICES, 2017a). Mapping prey hotspots and tern resilience Prey hotspots Areas with overlapping high abundances of multiple energy-rich prey species could act as prey hotspots for breeding terns and other seabirds, perhaps indicating which colonies are most resilient to fluctuations in prey populations. Such information could help refine the selection of tern colonies that are being considered for other conservation projects, such as expansion of the roseate tern. 24

27 Conversely, identification of areas with a single prey type or no abundant prey could highlight which colonies are most badly affected by food availability and would therefore benefit most from fisheries management or other interventions. Herring nursery grounds are shown in Figure 5 (Ellis et al., 2012). These areas are of interest because they contain high densities of juvenile herring which are generally an appropriate size for tern chicks, although large juveniles might be too big (a length of <17.5 cm was used to identify 0-group fish; Ellis et al., 2012). It should be noted that this included few surveys in shallow inshore waters; therefore inshore nursery grounds previously identified by Coull et al. (1998) are included in the maps by Ellis et al. (2012). Catch rates of sprat and sandeels around the UK and Ireland are presented in maps in the Fish Atlas by Heessen et al. (2015; not shown here). These maps were created using data from national and international research-vessel surveys carried out between 1977 and 2013, including the International Bottom Trawl Survey in the North Sea, the North Sea Beam Trawl Survey, the English Beam Trawl Survey, the Channel Groundfish Survey, the Celtic Seas Groundfish Surveys, the Norwegian Pandalus Trawl Survey, the Dutch Demersal Fish Survey, the Spanish Porcupine Bank Survey and the Baltic International Trawl Survey (Heessen et al., 2015). The data was largely taken from the ICES database DATRAS but also from other datasets that include areas not covered by international surveys (Heessen et al., 2015). Note the map of sandeel catch rates includes all Ammodytidae species. Although Heessen et al. (2015) also present a map of herring catch rates, this includes catch rates of adult herring. Adult herring in the North Sea tend to be between 20 and 30 cm in length (Heessen et al., 2005); in contrast, adult sprat rarely grow >16 cm in length (Whitehead, 1985). Thus, the map of herring catch rates by Heessen et al. (2015) would include individuals that are too large for tern chicks and is therefore not considered here. Figure 5. Herring (Clupea harengus) nursery areas around the UK and Ireland. Reprinted with permission from Ellis et al., (2012). Crown copyright,

28 Based on Figure 5 and the maps presented in Heessen et al. (2015), the distributions of high catch rates of sprat and juvenile herring appear to be quite aggregated. There are large herring nursery areas in the Moray Firth, Firth of Forth and the eastern and western Irish Sea. Substantial nursery areas also exist in the Minch and the Firth of Clyde, off western Scotland. The greatest aggregation of high sprat catch rates occur offshore of eastern England, to the south and east of the Dogger Bank (Heessen et al., 2015). However, high catch rates are also achieved in the Firth of Clyde, the eastern and western Irish Sea and patchily in the eastern and western Channel. Sandeels have a patchier distribution and fewer areas of high catch rates than sprat and herring. The largest areas with moderate to high catch rates of sandeels occur in offshore areas of the North Sea, suggesting that sandeels might be more accessible to terns breeding along the east coast of Britain than their counterparts in the Irish Sea and off the west coast of Scotland. In an attempt to summarise this information, a map of prey hotspots was created by overlaying the maps of sandeel and sprat catch rates from the Fish Atlas by Heessen et al. (2015) and herring nursery areas from Ellis et al. (2012). This is shown in Figure 6. It should be noted that this is not a precise account of the data, as it involved georeferencing the maps (thereby generating some error), selection of squares that did not always sit perfectly over the data points, and only shows catch rates that were deemed to be high for each species. However, this provides an approximate representation of areas that might be expected to hold an abundance of prey for breeding terns. A regular grid of 0.25 x 0.25 squares was created to approximately overlap the maps by Heessen et al. (2015). Squares with high catch rates of sprat (>1000 n/hour) or moderate-high catch rates of sandeels (>100 n/hour) were selected from the maps by Heessen et al (2015). The difference between the thresholds applied to the two prey types is due to sprat having a much higher proportion of squares with catch rates >100 n/hour; if squares with n/hour were included, most of the coast would be selected. In contrast, there are far fewer squares with catch rates of sandeels >100 n/hour. Therefore, different thresholds were applied in order to identify areas that have high catch rates for each prey. Following the same logic, squares with moderate-high maximum catch rates (>66,105) of juvenile herring were selected from the maps by Ellis et al. (2012). Only squares within 1 of the coast were included. Yellow squares are those with high catch rates of a single prey, orange squares have high catch rates of two prey types and the single red square has high catch rates of all three prey types. 26

29 Figure 6. Prey hotspots map showing approximate areas with high catch rates of three major prey types for terns (Ammodytidae sandeels, sprat and juvenile herring) based on data presented by Ellis et al. (2012) and Heessen et al. (2015). Red = high catches of 3 prey types; orange = high catches of 2 prey types; yellow = high catches of 1 prey type. Despite the general paucity of sandeels in the region compared to the North Sea, the Irish Sea has two relatively large hotspots of overlapping distributions of juvenile herring and sprat. The northeast coast of Ireland and the southeast coast of Northern Ireland has high to very high catch rates of both sprat and juvenile herring, explaining the preponderance of clupeids in the diet of terns at Rockabill (Hulsman et al., 2007; BirdWatch Ireland, 2008; Burke et al., 2016). In addition there exists a small area with moderate catch rates of sandeels ( n/hour) to the south of Rockabill, whereas the rest of the eastern coast of Ireland tends to have lower catch rates (Heessen et al., 2015). In the eastern Irish Sea, along the coasts of Cumbria and Lancashire, there is also substantial overlap between high catch rates of sprat and juvenile herring or sandeels. This area has very high catch rates of sprat, with 10,000-1,000,000 n/hour in many areas (Heessen et al., 2015). High and moderate-high catch rates of sandeels are restricted to the area of the coast lining the England-Wales border and offshore of the Lancashire coast, whereas high catch rates of juvenile herring are achieved further north. Some substantial tern colonies exist along the coasts of Cumbria and Lancashire, such as Seaforth Nature Reserve (172 Apparently Occupied Nests of common terns in 2016), Preston Docks (131 AONs of common terns in 2016), Foulney Island (54 Arctic tern pairs and 10 little tern pairs in 2016) and Hodbarrow (38 common tern pairs and 22 Sandwich tern pairs in 2016) (Roseate Tern LIFE Project, unpublished data). Although the diet of tern chicks at these colonies is unknown, based on the information presented in Figure 5 (Ellis et al., 2012) and the maps by Heessen et al. (2015) they are likely to contain a substantial proportion of clupeids. Offshore of northeast East Anglia are areas with overlapping high catch rates of sprat and sandeels, while East Yorkshire has offshore areas with high catch rates of sprat and juvenile herring or sandeels. However, some of these areas may be too far out to sea to be of much significance 27

30 for terns. Additional hotspots exist in the Moray Firth and Firth of Forth, which hold high abundances of sprat and juvenile herring and moderate abundances of sandeels in some offshore areas. A small area offshore of the Firth of Forth holds high catch rates of all three prey types. There are also prey cold spots where it appears that terns do not have access to large, diverse prey populations and therefore may be more sensitive to fluctuations in prey availability. Based on the prey hotspots map the central Channel is noticeably devoid of high catch rates of sandeels, sprat or herring. A locally abundant population of sprat is known to exist in Lyme Bay (ICES, 2013a), and the map of sprat catch rates in Heessen et al. (2015) shows that a few areas have moderate-high catch rates of up to n/hour. However, when looking over the maps it seems that the Channel has a considerably lower abundance of sandeels, sprat and juvenile herring compared to most other areas. There are also relatively low catch rates of sprat and low catch rates of sandeels and juvenile herring along the coasts of Durham and North Yorkshire. Although Coquet Island and the Farne Islands, off the Northumberland coast, appear to be in a prey cold spot, they sit just north of an area with a high abundance of sprat and within an area of a moderate-high catch rates of sprat. Additionally, moderate-high and high abundances of sandeels exist further offshore of the Northumberland coast. One area that was predicted to be a prey cold spot was the Shetland Islands, where seabirds are known to be particularly sensitive to sandeel abundance (Furness and Tasker, 2000). Although juvenile herring have been reported as absent or scarce around Shetland (Daan et al., 1990; Furness and Tasker, 2000), high catch rates have been observed, particularly to the southwest but also to the northeast and further offshore to the northwest (Figure 5). In the northeast these appear to overlap with a localised abundance of sandeels (Figure 6). However, excluding a few patchily distributed areas with moderate or moderatehigh abundances of sandeels and sprat, around most of the Shetlands the catch rates of these prey types are very low, and along the eastern coasts all three prey appear to be absent. It is important to note that the maps may miss small, localised areas with a high abundance of prey. For instance, based on the hotspots map juvenile herring appear to be largely absent from the waters around East Anglia, but young-of-the-year herring have been caught in relatively high numbers at Scroby Sands (Perrow et al., 2011). Figure 5 shows an inshore nursery area identified by Coull et al. (1998) ranging from eastern East Anglia to Kent, but this was not included in the hotspots map due to a lack of associated high catch rates. Therefore it is possible that a relatively high abundance of juvenile herring exists in the area but it was missed because there were insufficient inshore surveys to inform the maps by Ellis et al. (2012). This may also be the case in inshore areas between Northumberland and North Yorkshire, and along the Lincolnshire coast. Indeed, inshore surveys with an appropriate gear are recommended by Ellis et al. (2012) in order to improve knowledge of inshore herring nursery grounds. However, in each of these areas the surveys which did take place and were mapped by Ellis et al. (2012) showed very low catch rates, and therefore could not be included in the map of prey hotspots. A final note is that Figure 6 show prey hotspots based on survey data collected over the past 40 years. Populations of fish exhibit short and long-term shifts in distributions, and this variation is not captured by this map. Thus, this map cannot be used to make accurate predictions about the abundance of a certain prey species within an area in a specific year, as the fish may have temporarily moved out of the area. However, it highlights areas that are more or less likely to have a high abundance of prey, relative to other areas. Additionally, the distributions of each of these forage fish are sensitive to environmental conditions, and as such are likely to be shift in response to ongoing climate change. Thus, areas that currently have high abundances of multiple prey types may not in the future. 28

31 Resilience and vulnerability of tern colonies to prey shortage As terns generally behave as single prey loaders they are particularly sensitive to the energy content of individual prey, because a reduction in energy per prey item directly results in a reduction in energy per chick feed (Frederiksen et al., 2006; Österbrom et al. 2006, 2008; Engelhard et al., 2014). Colonies that can switch between alternative high energy prey species or target a mixture of high energy prey species would be expected to buffer their chicks from the effects of fluctuations in prey availability or body condition better than colonies that rely heavily on a single high-energy prey. Therefore, colonies with access to a high abundance of multiple high-energy prey species should experience fewer years of dietrelated breeding failures. Indeed, some tern colonies with access to only one high-energy prey species or a generally low abundance of high-energy prey have been shown to suffer lower breeding success than other colonies (Monaghan et al., 1989; Stienen and van Tienen, 1991; Stienen and Brenninkmeijer, 1998). Identifying areas where diet is unlikely to be a major limiting factor could be used to identify the colonies that are most resilient to fluctuations in prey populations. Based on this reasoning, colonies of common terns that have been active in the past 5 years with 50 AONs were plotted over the prey hotspots map (Figure 7). The data was collated and kindly provided by Chantal Macleod-Nolan (Roseate Tern LIFE Project). The common tern data is being used by the Roseate Tern LIFE Project to identify colonies that could support the expansion of the roseate tern in the UK and Republic of Ireland. Overlying the common tern information on the hotspots map could help locate colonies with an ample supply of food and therefore refine the selection of common tern colonies considered by the project. For instance, the colonies in the eastern and western Irish Sea have access to high abundances of multiple prey types. Rockabill, on the eastern coast of Ireland, already has a healthy population of roseate terns. However, there are no roseate tern colonies on the opposite side of the Irish Sea, on the west coast of England. Therefore, based on the availability of prey and the proximity of this area to a large source population of terns, colonies in this area may be suitable targets for management actions aimed at supporting the expansion of roseate terns. 29

32 Figure 7. Active common tern colonies with 50 apparently occupied nests (AON) in the past 5 years, overlying the prey hotspots map. While tern colonies in some areas are particularly resilient to fluctuations in prey populations, others are notably vulnerable. Based on the review of tern diet around the UK and Ireland, terns at Shetland are particularly vulnerable to changes in sandeel availability (Furness, 1982; Ewins, 1985; Monaghan et al., 1989; Suddaby and Ratcliffe, 1997; Monaghan et al., 1992; Furness and Tasker, 2000), although there appears to be an abundance of juvenile herring in some areas (Figure 5, Ellis et al., 2012). It may be that juvenile herring are not available to breeding terns at Shetland due to fine-scale spatial or temporal mismatch between herring nursery grounds and tern colonies. If sandeels continue to decline in the North Sea (ICES, 2013b; ICES, 2017a), terns at Shetland may find it increasingly difficult to find food. It is possible that other forage fish may move into the area in response to changes in environmental suitability. Sprat is currently more abundant in the southern North Sea (Daan et al., 1990; Harris et al., 2006; ICES, 2013a), and therefore a northwards shift of this forage fish in response to environmental change might benefit breeding terns in the Northern Isles. However, the likelihood of this is unknown, as the distribution of sprat (and other forage fish) is affected by many additional factors including oceanic currents (Corten, 1986; Daan et al., 1990; Baumann et al., 2006), the distribution and abundance of zooplankton prey (Hufnagl et al., 2013; ICES, 2013a) and interactions with other species (Rijnsdorp et al., 2010; ICES, 2013b). A brief review of possible climate change impacts, covering some of these additional factors, is given below. 30