
Leach's Storm-Petrel Distribution on Great Duck Island
Leach's Storm-Petrels (Hydrobates leucorhous) are Great Duck Island's most cryptic and most populous breeding seabird.
Please note: The analysis presented in this storymap, which was prepared as a class project for HS2020IM, is an early version of part of the analysis I conducted for my senior thesis. My thesis goes into much greater depth analyzing both the results of our 2022 petrel census and the history of petrel censuses on GDI, and can be accessed here: https://www.researchgate.net/publication/371200085_Comparing_Census_Methods_for_Leach's_Storm-Petrels_Hydrobates_leucorhous_A_Case_Study_from_Great_Duck_Island
Introduction
Great Duck Island (GDI) is a 91-hectare island lying 13 kilometers south of Mount Desert Island in the Gulf of Maine. GDI has a long history of human occupation, and has been farmed, grazed, and lived upon since the early 19th century. Today, approximately 85 hectares of the island are co-owned by TNC and the state of Maine and has been managed as a preserve since 1985. There is a small private inholding on the north end of the island, and the remaining five hectares are owned by the College of the Atlantic (COA). COA manages the Alice Eno Field Station out of the light station on the south end of the island, where students have conducted regular research on the ecology of the island since 1999 (Anderson 2018). For more information on the history and human use of the island, see Anderson 2018 and Stowe et al. (2021).
An adult Leach's Storm-Petrel on Great Duck Island
GDI is widely believed to be the largest colony of Leach’s Storm-Petrels (Hydrobates leucorhous) on the eastern seabird of the United States, although enormous colonies of these birds exist in Atlantic Canada, Newfoundland, and on the west coast of the U.S. Leach’s Storm-Petrels are small, pelagic seabirds in the tubenose order Procellariiformes. During the non-breeding season they have an extremely wide range encompassing most of the northern hemisphere and parts of the southern Atlantic. They are largely planktivorous, and feed by agitating the surface of the water with their feet in a manner typical to storm-petrels. Leach’s Storm-Petrels breed at colonies throughout the northern hemisphere (Pollet et al 2021). Like most procellariids they are long-lived, and usually return to the same colony and the same burrow for many years. Leach’s Storm-Petrels lay a single egg per year, and parental duties are shared evenly between the sexes. The breeding season is long; eggs are usually between mid-May and mid-June, although they can be laid as late as August (Chilelli 1999, Warham 1990). The incubation period lasts approximately 40 days, during which the breeding pair alternates incubation shifts that typically last between two and five days (Wilbur 1969). While away from the burrow, the non-incubating adult may go several hundred kilometers offshore to feed (Hedd et al. 2018). After hatching, chicks develop in the burrow for an additional 60-70 days before fledging. Once chicks achieve sufficient size to thermoregulate, their parents rarely remain in the burrow and typically return only at night to feed the chick (Chilelli 1999).
Leach’s Storm-Petrel burrows can be up to two meters long and often contain multiple twists and chambers (Chilelli 1999). The nesting chamber at the end of the burrow may be lined with a nest of vegetative material, though this varies by pair. On Great Duck Island, burrows are typically dug in loose soil and under cover objects such as fallen logs, roots, and boulders. Leach’s Storm-Petrels on GDI nest in both the red spruce forest and the island’s open meadows, though most research conducted on this colony has focused on the forest, where the birds are most abundant. Work by Kumagai (2019) and Stowe et al. (2021) indicates that the birds prefer the edge of the forest and blowdown habitats where woody debris is more common. For more information regarding the relationship between the Leach’s Storm-Petrel colony and the island’s varied habitat, please see Stowe et al. (2021).
Procellariids are one of the most rapidly-declining groups of birds in the world (Croxall et al. 2012). Concerning declines in Leach’s Storm-Petrel colonies have been recorded in Newfoundland, where the largest colonies exist (Wilhelm et al. 2019). They are listed as Vulnerable by the IUCN, and as a species of special concern by the state of Maine. These conservation considerations make population monitoring especially important. Leach’s Storm-Petrels are notoriously difficult to census (Ambagis 2002, Chilelli 1999). Their cryptic behavior, wide at-sea range, and burrowing habits mean that population surveys typically involve laborious burrow counts in densely-vegetated and inaccessible areas. The methods typically used for seabird censuses in the Gulf of Maine (surveys from a boat, aerial imagery, limited ground counts, etc.) are unsuitable for this species. The difficulty of obtaining accurate censuses is an obstacle to Leach’s Storm-Petrel conservation, and it is currently unknown whether or not the colony on Great Duck is in decline.
There have been ten population censuses on Great Duck Island for which information is available, dating back to 1973. All of these have been ground counts of burrows based on either transects or plots, though the details of methodology have varied widely. The results of these surveys have varied enormously—the lowest estimated 800 breeding pairs, the highest estimated approximately 27,000. The differences in methodology between these surveys makes them incomparable to each other and prevents inferences about trends in population; multiple surveyors have acknowledged that differences between their results and previous surveys likely did not reflect biological change (Folger and Wayne 1986, Stockwell 1995, Ambagis 2002). With a species that is so difficult to census, consistent methodology and an understanding of the factors that influence the results of population surveys are necessary to monitor the status of this colony.
During June - August of 2022, Levi Sheridan and I completed the most intensive survey on the Leach’s Storm-Petrel colony on Great Duck Island done to date. Our survey area covered approximately 7.8% of the available Leach’s Storm-Petrel habitat on the island and estimated a population of 8,461 +/- 1308 active burrows. We designed our survey methods to be comparable to Folger and Wayne’s (1986) survey, and adapted their stratified random sampling system of surveying subplots within a grid. This population estimate is comparable to many of those calculated by plot-based surveys in the past 50 years, but much lower than several recent transect surveys have suggested (Shannon 2018, Kumagai 2019). Shannon (2018) reported that the Leach’s Storm-Petrel colony on Great Duck has grown substantially since Folger and Wayne (1986) and Stockwell’s (1995) surveys. However, our results indicate that this may not be the case. In addition, our survey produced the first whole-island distribution map of this colony since 1986, which shows that the areas of the island where the colony is densest have changed in the past 40 years.
A Leach's Storm-Petrel burrow lined with debris.
Survey Methods
We based our survey methods off Folger and Wayne's (1986) study, which estimated a breeding population of 7040 pairs. They stratified the island into 50x50 meter grid squares, then randomly sampled a single 8.5x8.5 meter subplot within each square. We used the same method for stratification, and randomly sampled two 10x10 meter subplots within each grid square. This accounted for 8% of each grid square. Notably, Folger and Wayne (1986) built their grid off of magnetic north-south lines, using the tower on the south end of the island as an anchor point. We used the same anchor, but used true north-south to build our grid.
The 50 x 50 meter grid system. Zoom in to explore the grid squares, the labelling system, and the 10 x 10 meter subplots.
We used a Trimble GPS to locate each grid square, and used a random number generator to choose two subplots to sample. To locate each subplot, we stood at the nearest corner of the grid square and used a compass and tape measure to measure out each subplot edge. We used survey flags to mark the corners of each subplot and typically marked out one at a time, though when subplots were close to each other we occasionally marked out two at once. Grid squares were named with a number-letter system. Letters (A - R) corresponded to east-west columns of the grid, and numbers (00 - 38) corresponded to north-south rows. For example, a grid square in the northwest of the island had the ID C07, while a square in the southwest had an ID O33. Within each grid square, subplots were numbered 1 - 25. Use the embedded web map to explore the grid system, and zoom in on plot K20 to explore the 10x10 meter subplots.
We restricted our survey to the vegetated area of the island, and did not survey the rocky shoreline or the intertidal. Where grid squares were cut off by the shoreline, we adapted the number of subplots we surveyed. When more than 75% of the grid square was vegetated, we surveyed the usual two subplots. When 25 - 75% of the grid square was vegetated, we surveyed a single subplot. When less than 25% of the grid square was vegetated, we did not survey any subplots.
We did not survey subplots in the wetland in the center of the island, where the ground is frequently inundated with water and therefore unsuitable petrel habitat. To determine which grid squares to exclude, we walked along the boundaries of each 50x50 meter grid square within the wetland and determined whether any part of the square was outside the wetland. Squares that covered entirely wetland were not surveyed, while those along the boundary were surveyed normally. In addition, we only surveyed part of the private inholding on the north end of the island. We surveyed most of the forested part of the property, but avoided any grid squares within sight-lines of the house there. This was done by request of the property owners.
After setting out each subplot, we carefully counted petrel burrows within each 10x10 meter square. This involved combing the ground, checking under vegetation and around cover objects, and generally looking very closely for burrow entrances. Occasionally it was necessary to distinguish between ordinary holes in the ground and burrows by feel, and when we did this we generally considered holes that were wrist-deep or deeper to be burrows. With practice, it is very difficult to mistake a petrel burrow for anything else (Lesser 1977).
Within each subplot, we recorded the number of burrows present and whether or not each burrow showed signs of recent use (fresh digging, strong odor, or debris around the entrance that appeared intentionally arranged). For each subplot we also recorded any signs of predation, any human structures within the plot, and any other bird nests we found within the plot.
Visualizing the Survey Results
We surveyed a total of 310 50x meter grid squares and recorded data for 633 10x10 meter subplots. In addition, we surveyed all 25 possible subplots within three grid squares to explore burrow clustering on a smaller scale. In total, our survey covered approximately 7.8% of the island's petrel habitat. This is the most extensive survey done to date on Great Duck. Subplots were highlighted using a join between the grid map layer and our master data spreadsheet.
The location of every subplot we surveyed. Swipe to overlay the 50x50 meter grid.
The same table join allows us to connect a total burrow count to each subplot, which can be symbolized to show the relative petrel density in each subplot. It's immediately clear that this is a very clustered colony. There are large areas of the island without any petrels at all, and areas where they are very dense. Note that, even within the 50x50 grid squares where we counted all possible subplots, density can be very different within different subplots. The values shown in these subplots are direct counts--the actual number of burrows inside each subplot.
All the subplots we surveyed colored to show the total number of burrows in each. Swipe to overlay the 50x50 meter grid. Open the legend or click on a subplot to view its associated values.
Estimating Population
Population estimates that rely on sample areas (like this count) rely on a process called extrapolation. Extrapolation assumes that the values found in sample areas are equivalent to those that would be found, could you sample the entire area of interest. This allows us to calculate estimated (extrapolated) values for entire 50x50 meter grid squares. To do this, we assume that the average petrel density (i.e burrows per square meter) between the two subplots is the average density across the whole grid square. Put another way, we take the total number of burrows actually counted in each grid square and multiply it by the inverse of the percent of the grid square that we counted. For most grid squares, we counted two out of 25 possible subplots, or 8% of the grid square. This gives us an extrapolation factor of 12.5.
For example, in a grid square where we counted two subplots that contained eight and ten burrows, we would estimate that the grid square had a total population of 18 x (1/0.08), or 18 x 12.5, which gives an extrapolated value of 225 burrows. For shoreline plots, we have the advantage of knowing exactly how much of each grid square we sampled, which lets us be very precise with our extrapolations. However, extrapolation still relies on the assumption that our samples are representative of the overall area. Because petrel burrows are so clustered, even between subplots in the same grid square, we know that samples are far from perfectly representative. For that reason, we are much less certain of our population estimate for an individual grid square than we are of our estimate for the whole island.
Explore the map below to visualize the estimated burrow numbers per 50x50 meter grid square.
Estimated total burrows per 50x50 meter grid square. Click a square to view its associated value.
There is another important element of petrel colonies that complicates population estimates. Not every burrow corresponds to a pair of petrels. Some burrows on the island are empty; these may have belonged to birds that have died or moved to another burrow, or they may have been dug by overly-industrious birds at the beginning of the breeding season. Estimates of occupancy on GDI have ranged from around 33% (of burrows that actually contain a pair of petrels) to around 66%. How occupancy is estimated can be complicated, and every survey that has tried to account for occupancy has used different methods. Traditionally, burrowing seabirds are detected through grubbing, the process of manually exploring the burrow with a hand. However, this method is invasive and fails to detect activity in burrows that are too deep or convoluted to grub. Ambagis (2002) discussed methods of detecting occupancy, and concluded that grubbing was less effective than audio playbacks (where petrel vocalizations are played to elicit a vocal response from a bird in a burrow) or the use of a scope camera. However, grubbing may be the only reliable way to detect if a bird in a burrow is actually breeding (Stockwell 1995), since Leach's Storm-Petrels may occasionally take years off from breeding while still spending time in the burrow, or may "play house" for several years before actually attempting to breed.
A "stick test" where a petrel has had to move sticks aside to return to its burrow.
On GDI, we have often used stick tests and motion-activated camera traps as noninvasive methods of detecting occupancy. Stick tests involve popsicle sticks or twigs placed over a burrow entrance, so that the burrow's occupants must move the sticks aside to enter. Since incubation bouts typically last somewhere between 2-5 days, we typically check stick tests every day for several days to look for signs of use. Stick tests and camera traps are fairly non-conservative methods of estimating occupancy, compared to methods like grubbing, since they do not account for non-breeding or prospecting birds.
A petrel detected outside its burrow by a motion-activated infrared game camera.
In 2021, Lundy Stowe '22, Emily Rose Stringer '24, and I estimated burrow occupancy rates in 30 15 x 15 meter plots located around the island (Stowe 2021). Occupancy rates varied by plot, from 0% occupancy to 100% occupancy. Across the whole island, with a sample of 184 burrows, we estimated that 46% of burrows are active. This is, again, an estimate of the percent of burrows that are being actively used. It's not an estimate of the percent that contain an actual nest. See Stowe (2021) for more information on this occupancy estimate.
Occupancy is accounted for in a population estimate by simply multiplying the total number of occupied burrows by the proportion of occupancy, which is, in this case, 0.46. Shown below is the same map as above, adjusted to account for occupancy.
The distribution map of petrel burrows across the island, accounting for an occupancy estimate of 46%. Though I use the same color scheme throughout these maps, please note the difference between "active" burrows and total burrows.
The sum of the data shown in each of these grid squares would represent a whole-island population estimate (or, more accurately, a population estimate for the area we sampled, which doesn't include the wetland or the private inholding). However, because we know that our subplots are flawed representations of the overall grid squares, we can draw a more accurate estimate by considering the island as a whole.
Making a Final Population Estimate
The equation for this whole-island population estimate is, essentially, as follows:
Population (active burrows) = observed burrows * extrapolation factor * occupancy proportion
the extrapolation factor here is given by: sample area * ( 1 / area considered)
In this case, our "area considered" is the vegetated area of Great Duck Island, not including the wetland in the center of the island or the private inholding on the north end. Our sample area is the total area covered by our subplots, not including the extra subplots done in three grid squares (these would bias the estimate towards those densely-populated squares). GIS lets us calculate EXACTLY what area we covered and EXACTLY what area we're considering.
Our subplots, clipped to the permanent vegetation line, covered 52,592.8 square meters. The area of the island considered, which comes from vegetated area minus the wetland and the private inholding, covers an area of 665,852.9 square meters. This means we covered exactly 7.89% of the considered area, which gives us an extrapolation factor of 12.66.
Not including the extra subplots done in three grid squares, we counted 1485 burrows in our subplots. Our simple population estimate is, therefore:
1485 * 12.66 * 0.46 = 8,648 pairs of Leach's Storm-Petrels.
A histogram of the estimated number of burrows in each 50x50 meter grid square. 158 of our grid squares had no petrel burrows. The sum of the data shown in these histogram give a whole-island population estimate of ~8,600 active burrows.
This is a realistic estimate that's in line with several of the past surveys done on GDI. Great! However, this number is missing a measure of error, or a confidence interval. The usual method for generating a confidence interval would involve subsetting, or stratifying, our data, to compare how data are distributed between groups. These methods deal with sample distributions, which are usually represented by histograms, like the one on the right. One commonly-used method to do this would be to break the data down by habitat--for example, burrows in the forest might be one group, and burrows in the meadows might be another. However, most of these methods for estimating error assume that the data follow a roughly normal distribution. Our data is decidedly non-normal.
Histogram of potential whole-island estimates produced by non-parametric bootstrapping
For that reason, I decided to use a non-parametric bootstrapping method to calculate a confidence interval for our data. Bootstrapping is a method that uses sampling with replacement within an existing data pool to create potential sample distributions that can be compared. What I had the computer do was sample data for 310 plots, add and extrapolate them to get a population estimate, then repeat the process 10,000 time. This creates a distribution of possible population estimates. The mean value here, or the population estimate that the computer most frequently generated, is 8,461.
The normal distribution of these data also allow us to calculate a confidence interval, or, in this case, a credible interval. The difference between a confidence interval and a credible interval is worth looking up if you're very interested in the statistics here.
These methods give us a final population estimate of 8,461 +/- 1308 active burrows (95% credible interval). Just to note again, here, that active burrows means pairs of birds, but we don't know what proportion of active burrows contain breeding pairs. This is NOT an estimate of breeding pairs.
Comparison to Folger and Wayne (1986)
Folger and Wayne's (1986) original map of GDI with petrel densities drawn in. This survey was done as part of their large and baseline ecological inventory of GDI, which has been and continues to be an important resource for research on the island.
Our survey produced the first colony-wide distribution map of Leach's Storm-Petrels on GDI since Folger and Wayne's 1985 survey. We designed our methods to be directly comparable to theirs, although we used different methods to account for occupancy. If they had used the same estimate of occupancy that we did (46%), they would have estimated 8117.62 active burrows. This is strikingly similar to our estimate and well within our credible interval. They actually reported 11,734 active burrows, having estimated approximately 66% occupancy. While occupancy likely does vary by year, the difference in methodology between our two studies make it impossible to detect whether the rate has actually changed. Folger and Wayne predicted that the Leach's Storm-Petrel population on GDI would rise as the impacts of sheep grazing faded from the island. Today, the population is not significantly different. Unfortunately, the surveys that have been done in the interim years have tended to use very different methods and aren't comparable to each other, so we can't say whether the population has changed and then changed back, or stayed stable over 40 years.
So, we can't detect a change in total population, but we can visualize how the colony's distribution has changed. This required digitizing Folger and Wayne's map, which was drawn over digital imagery. Unfortunately, because their grid used magnetic north and ours used true north, the grid squares aren't directly comparable. However, we can still see spatial changes in the colony. .
A comparison of Folger and Wayne's 1985 petrel density map and the 2022 density map. The 1985 grid appears slightly distorted--the map used for that grid was based off of aerial imagery, while the 2022 grid was built on GPS coordinates, so the georeferencing is not the same. Also, please note that the color grading used here is different than the grading used on the density map above. The way Folger and Wayne reported their data restricts this symbolization.
Looking at the change between 1985 and today reveals several of places where petrel density has changed. The population on the north end has increased, while the population south of the wetland has cleared out some. The west side of the island continues to be a stronghold, but the clusters there are seem to have spread out a bit. There are more petrels present in the south end around the gull colony, and fewer in the middle of the island. Understanding why these changes have occured would be an interesting avenue of future study.
Conclusion
The Leach's Storm-Petrel population on Great Duck Island does not appear to be significantly different than it was in 1985. However, the distribution of these birds around the island has changed, particularly on the north end. A history of inconsistent count methods prevents us from tracking changes in the colony's population. Understanding the distribution of this colony can help inform future surveys to produce more precise and more comparable counts in the future.
Uncertainty around occupancy rates--the proportion of burrows that are active, and the proportion of those that actually breed--is an unfortunate reality of these kinds of estimates. Occupancy likely varies by year and certainly varies spatially across the island. It is important to understand the difference between an estimate of active burrows and breeding pairs. Seabird surveys and atlases are often most interested in estimates of breeding pairs, and these two categories are often used interchangeably without actually being the same. Leach's Storm-Petrels are long-lived birds that are typically loyal to their colony and burrow. They can breed many times in their lives, and may take years off of breeding despite being active members of the breeding population. Young birds may also spend multiple years prospecting, digging burrows, and building pair bonds before breeding. Population estimates that attempt to only include burrows that actually contain an egg or chick may neglect these members of the colony. When planning a survey, it's worth considering which value--the number of active burrows or the number of breeding pairs--is actually of greatest interest.
Works Cited
Ambagis, J. 2002. Census and Monitoring Techniques for Leach’s Storm Petrel (Oceanodroma leucorhoa). Thesis submitted in partial fulfillment of the requirements for an MPhil degree, College of the Atlantic, Bar Harbor, ME.
Anderson, J. 2018. Great Duck Island: A Preliminary Report 1999-2018. Report to The Nature Conservancy, Brunswick, Maine. 66pp.
Chilelli, M. 1999. Leach’s Storm-Petrel Assessment. Maine Department of Inland Fisheries and Wildlife: Wildlife Resource Assessments. 42pp.
Croxall, J.P., Butchart, S.H., Lascelles, B., Stattersfield, A.J., Sullivan, B., Symes, A. and Taylor, P., 2012. Seabird conservation status, threats and priority actions: a global assessment. Bird Conservation International, 22(1), pp.1-34.
Folger, D. and Wayne, P. 1984. A Biological Inventory of Great Duck Island. College of the Atlantic, unpublished report to Maine Nature Conservancy. 97pp
Hedd, A., Pollet, I.L., Mauck, R.A., Burke, C.M., Mallory, M.L., McFarlane Tranquilla, L.A., Montevecchi, W.A., Robertson, G.J., Ronconi, R.A., Shutler, D., Wilhelm, S.I. and Burgess, N.M. 2018. Foraging areas, offshore habitat use, and colony overlap by incubating Leach’s storm-petrels in the Northwest Atlantic. Plos One 13(5):p.e0194389.
Kumagai, A. 2019. Effects of Vegetation on Leach’s Storm-Petrels Nest Site Preference on Great Duck Island, Maine. College of the Atlantic. Poster Presented to the Waterbirds Society.
Pollet, I. L., A. L. Bond, A. Hedd, C. E. Huntington, R. G. Butler, and R. Mauck. 2021. Leach's Storm-Petrel (Hydrobates leucorhous). Birds of the World. Cornell Lab of Ornithology, Ithaca, NY, USA.
Shannon, P. 2018. Population Assessment of Leach's Storm-Petrel in Maine. Report to the Maine Outdoor Heritage Fund. 8pp.
Stockwell, K. 1995. Population census of Leach's Storm- Petrels, Oceanodroma leucorhoa, Great Duck Island, Hancock County, Maine. Report to the Maine Chapter of the Nature Conservancy. 4pp.
Stowe, L., Morrel, H., and Gnam, E. 2022. Leach's Storm-Petrel Research and Conservation on Great Duck Island. Unpublished Documents, COA Great Duck Files. 27pp.
Wilbur, H.M., 1969. The breeding biology of Leach's Petrel, Oceanodroma leucorhoa. The Auk 86(3):433-442.
Wilhelm, S.I., Hedd, A., Robertson, G. J., Mailhiot, J., Regular, P. M., Ryan, P. C. and Elliot, R.D. 2019. The World’s Largest Colony of Leach’s Storm-Petrels (Hydrobates leucorhous) Has Declined. Bird Conservation International 30:40-57
Acknowledgements
The data, analysis, and research presented in this storymap come from and make up a large part of my senior project. As with any project of that size, I have a long list of people to thank. All of the research done on Great Duck Island, especially with Leach's Storm-Petrels, is part of a legacy of student work on the island, and we all stand on each others' shoulders. I can't thank, in name, every student whose work I have benefitted from, but I am deeply grateful to all the generations of past Ducklings, to the college, and to Alice Eno for the opportunity to do the work that I've done on the island.
No part of this project would have been possible without the hard work and unflinching dedication of Levi Sheridan. I am deeply grateful to have him as half of my field team.
The GDI 2022 Field Crew
I want to thank all the members of the 2022 Great Duck crew for their general support, assistance, and good company: Wriley Hodge, Asher Panikian, Jennifer McNamara, Hannah Gower-Fox, Rosie Chater, and Marina Schnell. In addition, parts of this project have built directly on work I did in the 2021 season alongside Lundy Stowe, Emily Rose Stringer, and Finley O'Connor. I am also grateful to Toby Stephenson and the crew of the RV Osprey for the work they do to make our field seasons possible. I am also grateful to Rich and Andra Burofsky for granting us permission to count on parts of their property.
Over the course of this project I've recieved invaluable advice and support from numerous people, and I'd especially like to thank Liam Taylor, Kate Shlepr, Dave Folger, Aya Kumagai, Eben Sypitkowski, and Chris Petersen.
I am constantly grateful for the mentorship and guidance of John Anderson, who advised this project and made it possible.
Throughout the course of my work on Great Duck Island, I've recieved financial support from Downeast Audubon's Sal Rooney memorial fund, Maine Sea Grant, the Maine Space Grant Consortium, the William H. Drury Research Fund, and the Waterbirds Society Student Travel Fund. I am deeply grateful for their support.
A Leach's Storm-Petrel chick
Photo Gallery