Quantifying gull predation in a declining Leach’s Storm-petrel ( Hydrobates leucorhous ) colony

. The effect of gull predation on sympatric seabirds has garnered much attention and management action in recent decades. In Witless Bay, Newfoundland, Canada, gulls depredate significant numbers of Leach’s Storm-petrels ( Hydrobates leucorhous ) annually. We quantified this predation on Gull Island in Witless Bay, and its effects on the storm-petrel population, by estimating the annual gull predation rate using strip transects to count storm-petrel carcasses and predicting storm-petrels’ population growth rate by repeating an island-wide breeding census. Using methods that account for island topography, we found that the Leach’s Storm-petrel breeding population on Gull Island declined to roughly 180,000 pairs in 2012 (95% CI: 130,000–230,000), a decrease of 6% per year since the last census in 2001 (352,000 pairs). Based on carcass counts, gulls, mostly American Herring Gulls ( Larus argentatus smithsonianus ), depredated 118,000–143,000 Leach’s Storm-petrels in 2012. Studies of storm-petrel recruitment, the contribution of the large non-breeding component of the population to gulls’ diets, and the consequences of gulls’ storm-petrel diet on the gulls themselves are needed to better predict the trajectory of both species into the future


INTRODUCTION
The conservation and management of species is complex when there are competing conservation interests, and actions to benefit one species may be detrimental to another, necessitating an ecosystem-based approach (Williams et al. 2011). Conservation actions that focus on one species, or on individual species without considering species interactions, are unlikely to be broadly beneficial (Soulé et al. 2005). Such is the case when both predator and prey species are in decline (Chadès et al. 2012).
The Witless Bay Seabird Ecological Reserve, located off the east coast of Newfoundland and Labrador, Canada, contains two important seabird colonies where American Herring Gulls (Larus argentatus smithsonianus) and Leach's Storm-petrels (Hydrobates leucorhous) nest sympatrically: Great Island (47°11'N, 52°46'W) and Gull Island (47°15'N, 52º46'W). Atlantic Herring Gull populations on the two islands have decreased in recent years, from a high of 3852 pairs in 1979 (Cairns and Verspoor 1980, unpublished report) to 1608 pairs in 2011 on Gull Island, and from a high of 2771 pairs in 1979 (Cairns 1979) to 358 pairs in 2012 on Great Island . In spite of these declines, gulls have expanded their nesting habitat into coniferous forests where storm-petrels breed (Robertson et al. 2001). Great Island supports one of the largest Leach's Storm-petrel colonies in the western North Atlantic, but surveys conducted in 2011 revealed that the population had declined from 300,000 pairs in 1979 to 134,000 pairs (Wilhelm et al. 2015). Gull Island supports a slightly larger storm-petrel population, with an estimated 350,000 pairs, but the most recent survey dates back to 2001 (Robertson et al. 2002). Together, the islands of Witless Bay are among the most important for breeding Leach's Storm-petrels globally and are, therefore, recognized as an Important Bird and Biodiversity Area (Birdlife International 2022a, Pollet et al. 2021). Declines in Leach's Storm-petrels have been sufficient to result in the species being listed as Vulnerable on the IUCN Red List (Birdife International 2022), and Threatened in Canada (COSEWIC 2020).
The western North Atlantic Ocean has experienced considerable natural and anthropogenic changes in the last 40 years. A moratorium on groundfish fisheries in 1992 significantly reduced the availability of fisheries' discards to seabirds and levels of gillnet bycatch (Stenhouse and Montevecchi 1999, Regular et al. 2013. Shifts in the oceanographic regime also delayed the timing of capelin (Mallotus villosus) spawning, a key component of many seabirds' diet, resulting in complex species interactions (Regehr and Rodway 1999, Stenhouse and Montevecchi 1999, Massaro et al. 2000, Carscadden et al. 2001, Baillie and Jones 2003, Davoren and Montevecchi 2003, Bond 2016. These changes were expected to have resulted in increased predation on Leach's Storm-petrels by gulls, as they sought to bridge the nutritional gap between chicks hatching, and capelin spawning Montevecchi 1999, Stenhouse et al. 2000). Though oceanographic and climatic conditions are strong drivers of Leach's Storm-petrel populations, anthropogenic influences are increasingly important (Duda et al. 2022). Initial estimates in 1997 found that 49,000 storm-petrels were depredated annually on Great Island, raising concerns for the storm-petrel population's persistence (Stenhouse et al. 2000). However, this study also predicted that predation on storm-petrels would decrease with declining gull numbers (Stenhouse et al. 2000).
Large gulls have been implicated in the population declines of sympatric waterbirds, though the evidence suggests they likely have little long-term effects on populations (Oro and Martínez-Albraín 2007). Storm-petrels, in particular, experience high predation pressure from gulls (Stenhouse et al. 2000, Oro et al. 2005, Matias and Catry 2010, Hoeg et al. 2021. Storm-petrels can have high rates of pre-breeding dispersal (Warham 1990, Brooke 2004, which means that quantifying the effects of gull predation on storm-petrel populations is confounded by the abundance of young, pre-breeding, and prospecting storm-petrels around breeding colonies but originating elsewhere (Bicknell et al. 2014). Nonetheless, estimating the number of storm-petrels killed by predators at colonies remains important as it is one of the few direct mortality sources that can be quantified for this species.
Our objectives were to update the status of the Leach's Stormpetrel population on Gull Island by conducting a survey accounting for habitat topography. A Geographic Information System (GIS) approach (Wilhelm et al. 2015) was used to estimate gull predation on storm-petrels on Gull Island by conducting strip transects for storm-petrel remains and estimating the total number of storm-petrels killed by gulls by extrapolating our counts to the area occupied by storm-petrels, again accounting for topography, and determining how gull predation of stormpetrels varied within years.

Colony mapping
On Gull Island, storm-petrels are found breeding in the dense forest (primarily spruce and fir) in the interior of the island covering a flat and broad valley. Atlantic Puffins (Fratercula arctica) occupy the grass habitat surrounding the forest but are expanding their breeding range into the outer edge of the forest (COSEWIC 2020). The extent of the area occupied by stormpetrels and puffins in 2012 was determined by using a highresolution digital aerial photograph taken in 2008 and purchased from Surveys and Mapping Division, Department of Environment and Conservation, Government of Newfoundland and Labrador. This image was georeferenced in ArcGIS 10.0 (Esri, Redlands, California, USA) and served as the basis to delineate the forested area occupied by storm-petrels and grass habitat occupied by puffins. The delineation between the puffinand storm-petrel-occupied forest was further refined and groundtruthed by walking along the edge of the forest and identifying the border where puffin burrows ended and storm-petrel burrows began, while holding a handheld GPS to acquire location coordinates at 10 m intervals (accuracy of ≤ 5 m; Fig. 1).

Occupied area, burrow density and occupancy
A topographic map of Gull Island at a scale of 1:943 with 3.048 m (10') contour intervals was digitized in ArcGIS 10.0. A threedimensional (3D) island was subsequently extruded in ArcScene based on the elevation image. The digitized boundaries of the storm-petrel habitat for 2012 were subsequently draped over the 3D island allowing the 3D slope-corrected occupied area to be calculated, as detailed in Wilhelm et al. (2015).
We overlaid an island-wide 100 × 100 m grid over the georeferenced aerial photograph and retrieved the coordinates for intersecting lines that fell within what would be described as storm-petrel habitat (n = 44). The grid was placed randomly over the island. Between 3 and 19 July 2012, coinciding with the stormpetrel chick-rearing period, each of these points were visited to confirm that the habitat was suitable for nesting storm-petrels, and if so, a 3 × 3 m plot was established to assess burrow density and occupancy. Within each plot containing storm-petrel holes, observers inserted their arm down each hole and tunnel to assess its contents and assigned each hole to one of the following categories: 1) extra entrance or entrance to another burrow, 2) too short to be a burrow (< 30 cm), or 3) a suitable burrow. For this final category, the contents of the burrow was recorded (empty, adult, adult and egg, egg only, adult and chick, chick only, or could not be determined).
Burrow density was calculated for each plot by dividing the number of suitable burrows present by the area sampled (9 m²). Burrow occupancy was derived by calculating the proportion of occupied burrows (i.e., burrows containing an adult and/or an egg or chick) in relation to all burrows where contents were assessed (i.e., occupied burrows and empty burrows but excluding burrows where the contents could not be determined). Island-wide burrow density and occupancy was estimated as the mean and variance of the plot-level burrow densities and occupancies.

Statistical analyses
The population estimate of Leach's Storm-petrels breeding on Gull Island in 2012 was calculated as the product of island-wide burrow density and occupancy rate and total occupied area as determined using the 3D analysis tool in ArcGIS 10.0. Burrow density and occupancy rates are presented as mean ± 1 SE. Population estimates are presented as the mean, while standard errors of the mean, and associated 95% confidence interval (C. I.), were based on calculations from the standard formula for calculating the variance of a product (ignoring covariance).

Quantifying gull predation on storm-petrels
Methods followed those of Stenhouse et al. (2000). We established three 2 m-wide transects on Gull Island in 2012 and 2013. The first was 407 m long and went from the southern tip of the island ("The Finger") through mixed rocky habitat (49 m), meadow habitat (215 m), and puffin habitat (77 m), all of which has sympatric breeding gulls , terminating in coniferous forest (66 m; Fig. 1, A), where gulls do not breed. The second transect followed an established path used by researchers for 172 m through forest (B), and the third transect through dense forest for 49 m, away from any regularly used path (C). Transects B and C did not cross gull breeding habitat. Habitats were classified at the beginning of the season and differ somewhat from those presented in the habitat mapping of this study because of the difficulty in separating puffin habitat from meadow habitat in aerial photographs. At the beginning of the 2012 and 2013 storm-petrel breeding seasons, we walked the three transects and removed any evidence of storm-petrel predation; regular surveys began one week later. Surveys were conducted from 12 May to 27 July in 2012 and from 16 May to 21 August in 2013. Every 7 days (on two occasions the interval was increased to 14 days), we walked the entire length of each transect to record evidence of storm-petrel predation. This included carcasses, wings, or gull pellets containing storm-petrel remains. Once identified, the evidence was removed to prevent repeated counting. Given the small area searched on either side of the transect (1 m), and the high visibility of storm-petrel remains, we assumed all petrel remains were found (i.e., that the probability of detection was 1). Using similar methods, Hoeg et al. (2021) found that by using two observers, 89% of storm-petrel remains were found by one observer, so our assumption of complete detection may underestimate predation levels. Leach's Storm-petrels typically breed for the first time when 5 years old and rarely return to their natal colony , resulting in a large pool of prospecting individuals who investigate several breeding colonies, often covering thousands of km, throughout the summer for future breeding attempts (Bicknell et al. 2014). We could not distinguish the remains between these prospecting individuals and local breeding storm-petrels, or local breeders who had failed.
We recalculated the number of storm-petrels killed by gulls in 1997 for nearby Great Island (Stenhouse et al. 2000) by using the occupied area estimated for 1994, namely 154,804 m² of forest and 112,875 m² of open habitat (Wilhelm et al. 2015; note that storm-petrels nest in both habitats on Great Island), but which considered topography using the same GIS approach as this study. This allowed both estimates of mortality to be directly comparable. For 1997 on Great Island (using transect data presented in Stenhouse et al. 2000) and 2012-2013 on Gull Island, the number of storm-petrel carcasses found on transects was converted into a rate per m², and then scaled up based on total area of each habitat to provide an estimate of the number of storm-petrels depredated by gulls on each sampling occasion, the sum of which represented the gulls' entire breeding season (May-August).
For both time periods, we examined temporal trends in stormpetrel predation through their breeding season by plotting the

Leach's Storm-petrel breeding population on Gull Island
In 2012, we recorded the lowest density of storm-petrel burrows on Gull Island, including a decrease of 44% compared with the survey in 2001 (Table 1). The two-dimensional area occupied by storm-petrels in 2012 was estimated at 452,623 m². Using the 3D analysis which considers sloped habitat, the occupied area was estimated at 462,633 m², increasing the estimated occupied area by 2%. In 2012, storm-petrel burrow density averaged 0.604 ± 0.064 burrows/m² (n = 44 plots) with an average burrow occupancy rate of 0.598 ± 0.051 (n = 40 plots as 4 plots did not have any burrows), yielding an average of 0.389 ± 0.055 occupied burrows/m² (n = 44 plots). Extrapolating the occupied burrow density across the 3D occupied area of 462,633 m², we estimate that 179,742 ± 25,340 (130,078-229,407; 95% CI) pairs of Leach's Storm-petrels were breeding on Gull Island in 2012 (Table 1).

Gull predation
In 2012 on Gull Island, we located evidence of 38 storm-petrels killed in forest habitat, 39 in meadow habitat, and one each in rocky and puffin habitat. This extrapolated to 118, 622 kills in forest, 22,942 in meadow, and 568 in each of puffin and rocky habitat for a total of 142,701 storm-petrels killed. In 2013, we located evidence of 52 storm-petrels killed in forest habitat, 19 in meadow habitat, two in puffin habitat, and one in rocky habitat on Gull Island. This extrapolated to 100,888 kills in forest, 12,352 in meadow, 1137 in puffin habitat, and 568 in rocky habitat for a total of 114,945 storm-petrels. Stenhouse et al. (2000) originally estimated that 12,653 stormpetrels were killed in open/meadow habitat, and 36 536 in forested habitat on Great Island in 1997. Extrapolating the transect data to 3-dimensional habitat areas (Wilhelm et al. 2015) gave estimates of 19,649 kills in meadow habitat, and 55,058 in forested habitat, for a total of 74,707 storm-petrels killed on Great Island from 18 May to 21 August 1997. This is a 52% increase from the original estimate (49,189 storm-petrels; Stenhouse et al. 2000). Predation decreased over the storm-petrel breeding season in both 1997 and 2012/2013 (t = -3.88, p = 0.0001, β ± SE: -0.0062 ± 0.0016 birds/ m²/day; Fig. 2).

Leach's Storm-petrel population
Burrow density was also low in the 1980s on Gull Island, and the reasons for the increase in 2001 are unknown. Storm-petrel burrows are small, and without being regularly occupied, are unlikely to persist on Gull Island more than two or three breeding seasons (A. Hedd, personal communication). Decreases in burrow density could indicate population declines over time, or a decrease in prospecting activity by pre-breeders. Unoccupied burrows likely represent lack of use during the current or recent breeding season, as burrows are still viable and simply unused. There has been a steady decline in burrow occupancy on Gull Island since the 1980s and in other Canadian colonies (d' Entremont et al. 2020) which, in other burrow-nesting species, has been attributed in part to predation (Cuthbert et al. 2013).
We estimated that the Leach's Storm-petrel population on Gull Island has decreased by 6% per year between 2001 and 2012. As long-lived birds with low annual fecundity, adult survival is, therefore, the most influential demographic rate in determining population trajectories (Morris and Doak 2002). Leach's Stormpetrels throughout eastern Canada have low apparent adult survival (approximately 80% on Bon Portage Island, Nova Scotia), and a potential source of mortality is predation from sympatric gulls (Fife et al. 2015, Wilhelm et al. 2019, which is mirrored in our study in Witless Bay. Conditions on migratory routes and wintering areas in the central Atlantic Ocean, such as diet and physiology, could influence survival (Pollet et al. 2014, Fairhurst et al. 2015, Pollet et al. 2019, as could contaminants, such as mercury (Bond andDiamond 2009, Pollet et al. 2017). While at their breeding sites, other known threats aside from predation include attraction to artificial light (Gjerdrum et al. 2021, Wilhelm et al. 2021) and interactions with offshore oil and gas facilities (Ellis et al. 2013. This combination of threats and population declines resulted in the species being up-listed to Vulnerable on the IUCN Red List (Birdlife International 2022b) and the eastern Canadian population being assessed as Threatened by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC 2020) .

Gull predation
By accounting for slope, we increased the estimate of stormpetrels depredated by gulls on Great Island by 52%, namely from 49,000 (Stenhouse et al. 2000) to 75,000. This is more in line with the estimates from nearby Gull Island in 2012 and 2013 (118,000-143,000 storm-petrels depredated) and highlights the importance of accounting for topography when extrapolating based on surface area. However, the conclusions in Stenhouse et al. (2000) regarding the expected impact of this predation remain valid, as the population size on Great Island was also underestimated in 1997 (Wilhelm et al. 2015). The majority of these depredated birds will be prospecting non-breeders. Even if the entire 6% annual decline in breeding pairs (equivalent to 21,500 individuals) was attributable to gull predation, it represents < 20% of the annual estimated predation. The recruitment and immigration dynamics of Leach's Storm-petrels are complex, with few birds returning to their natal colony to breed ) and large numbers of birds from distant colonies prospecting (Bicknell et al. 2014).
Most of the predation was attributed to American Herring Gulls. Great Black-backed Gulls (Larus marinus) breed sympatrically but rarely consume storm-petrels (Veitch et al. 2016). Based on populations of 88 Great Black-backed Gulls pairs in 2001 on Gull Island (Robertson et al. 2001), and 2054 Leach's Storm-petrels depredated (i.e., 11.7 storm-petrels depredated per Great Blackbacked Gull) (Veitch et al. 2016), and extrapolating to the 2012 population (32 pairs; Bond et al. 2016), Great Black-backed Gulls likely consumed only 750-1000 Leach's Storm-petrels (i.e., < 1% of the estimated total predation) annually. Non-breeding and immature gulls are unlikely to contribute significantly to predation at the colony, as nesting gulls are highly territorial and likely to drive off intruders (Pierotti and Annett 1994), and immature gulls are rare at breeding colonies (authors personal observation).
In 2000 (the nearest census year to Stenhouse et al. 2000), the American Herring Gull population on Great Island was 1640 pairs , which equates to a predation rate of 22.3 storm-petrels/gull. This is nearly half of what we observed in 2012-2013 (44.3 storm-petrels/gull). These results suggest that despite continued population declines of gulls between 2000 and 2012, predation pressures on storm-petrels are not subsiding. This could reflect decreasing availability of gulls' other prey, behavioral shifts, or other factors which merit further investigation.
Most storm-petrel depredation events occurred in forest habitat, which is where their burrows are located. Gulls have slowly been encroaching into this habitat in Witless Bay (Robertson et al. 2001), and though the gull population has declined overall, gulls breeding in forest habitat have declined more slowly . Because nesting habitat is related to gulls' diet and breeding success (Pierotti 1982, Pierotti andAnnett 1991), there must be some advantage for gulls of nesting in forest habitat. The readily available supply of both breeding and prospecting storm-petrels may be one such advantage.
Our findings further support the previous observation that despite high energetic demands associated with chick rearing, predation on storm-petrels decreased over the gull breeding season (Stenhouse and Montevecchi 1999). For both study periods, the highest predation rates were observed in May and June and dropped suddenly by early July, coinciding with peak inshore capelin spawning, which during the course of these two studies occurred in early to mid-July Montevecchi 1999, Fitzsimmons et al. 2017). In addition to switching prey from storm-petrels to capelin, gulls may also take advantage of the presence of other seabird chicks, which begin to hatch in late June and early July (e.g., Atlantic Puffin, Common Murre, Blacklegged Kittiwake), and can form significant parts of gulls' diets (Bond 2016).
Herring Gulls are significant predators of Leach's Storm-petrels in Witless Bay, Newfoundland, and may be partly responsible for the significant decrease in the breeding storm-petrel population. However, it's important to note that similar population declines have been observed in the world's largest Leach's Storm-petrel colony on Baccalieu Island, despite very few gulls breeding on the island (Wilhelm et al. 2019). The annual predation of 118,000-143,000 storm-petrels, though large, is biologically plausible given the large population of storm-petrels in the northwest Atlantic and the significant energetic requirements of the relatively large gull breeding populations in Witless Bay (Pierotti 1982, Pierotti andAnnett 1991). There is also considerable movement of prospecting individuals and presumed failed inexperienced breeders among colonies between Newfoundland, Iceland, and the United Kingdom (Bicknell et al. 2014). Recruitment into the breeding storm-petrel population, and the ecotoxicological consequences of this predator-prey system where storm-petrels have elevated Hg concentrations (Bond and Diamond 2009), needs further study. Investigation into spatial heterogeneity is also warranted, both estimates of stormpetrel population trends and levels of gull predation. Current management strategies aimed at restoring seabird populations by controlling gulls may not be as immediately effective unless they can increase recruitment in gulls' prey species. Range-wide decreases in gull populations (Anderson et al. 2016) mean current management practices may not be appropriate in some cases, and gull-seabird dynamics may be shifting. An ecosystem-based approach is likely the only solution for the successful conservation of both gulls and their storm-petrel prey (Slocombe 1993), which includes mitigating specific threats and understanding the synergistic relationship between gulls, storm-petrels, anthropogenic subsidies, and the state of the broader marine environment.